+ studio updates

1. New home. keiki weekly is no longer at keikistudio.substack. From this issue on, it lives on product.blog. Same writing, same cadence, same build-in-public stance. We just stopped running two publications when one will do. Subscribers were migrated. Nothing else changes.

2. Iberia is calling. A handful of investors from the peninsula reached out over the last two weeks to talk. We took the calls. Nothing concrete to share yet, but the questions an early investor conversation forces you to answer are useful in the same way this newsletter is useful. They make you put words to things you would otherwise leave implicit. If anything moves, it will land here first.

+ product updates

Weekly #13 framed Corexi as a continuous UX layer for AI-native products. The thesis: the products of the next decade will be written largely by AI agents, and a one-shot scan tool isn’t enough. UX needs to be a layer the agent lives inside, not a report it reads after the fact.

These two weeks were about making that thesis concrete. We didn’t ship a new tab. We shipped a new location.

Corexi is now an MCP server.

It runs inside Cursor, Claude Code, Windsurf, Replit, and VS Code through the Model Context Protocol, the open standard for how IDE-resident agents talk to external tools. Eight tools the agent can call without leaving the editor: pull a previous scan, read the Corexi ruleset, review a snippet or a screenshot before it ships, trigger a fresh scan from the IDE.

Auth is real. Hashed API keys, rate limits, monthly quotas. Signup drops you at your first key and an IDE config snippet in three screens.

The result is one engine serving two personas. At build time, the agent asks for a review while writing the component and gets back a fix-ready snippet that already uses your design tokens. Post-deploy, it pulls findings from yesterday’s scan and patches them in your repo.

Landing at corexi.com/mcp. Setup at /docs/mcp.

The engine got more trustworthy.

Five things landed inside the engine itself.

Every LLM finding is now cross-checked against axe-core’s deterministic accessibility rules. You can tell at a glance which findings have a hard reference and which are model judgement calls.

Design tokens are now ingested and injected into every prompt, and a second pass walks the generated fix_code afterward, finds any hardcoded hex or px that falls inside a perceptual color and length threshold, and rewrites it in place. Once a team wires their tokens, switching cost rises sharply.

We built an eval set. Curated fixtures with hand-labeled ground truth, a fuzzy matcher, a CI guard that fails the build if recall or precision drops more than five points. We can now say something concrete about how the engine performs instead of relying on vibes.

Multi-state capture. The scanner now triggers focus, modal, and form-validation states on each page, screenshots them, runs axe on the new DOM, and analyzes them as separate captures. Catches a whole class of bugs that static screenshots miss.

A real Lighthouse audit runs on every scan. Performance, accessibility, best practices, SEO, plus Core Web Vitals. Surfaces in the report and through the MCP tools.

Visibility, both sides.

Admin panel for us, dashboard for users, telemetry table underneath. Every MCP tool call is recorded with tool name, status, duration, and detected IDE. Also the foundation for usage-based billing later in the year.

Where this leaves us.

Corexi is no longer a scanner with a dashboard. It’s a verified engine, a continuous review surface inside the IDEs where AI-native code actually gets written, and a per-user telemetry stream that proves it’s being used.

Two personas. One layer. Wherever developers and agents write code.

Next up: an interactive no-login showcase at /mcp/examples, per-state badges in the report UI, and a Cursor marketplace submission so install becomes one click. Pricing is queued behind these.

A Note on Location

There’s a difference between building a tool and choosing where it lives. The first weeks of Corexi were about building the tool. These two were about choosing the location. For AI-native products, location matters more than features. A scanner you have to remember to open is a scanner you forget to open. A scanner the agent calls inside the IDE is a scanner the agent calls all day.

If you’re building an AI-native product and want UX guidance during the build instead of after it, the MCP server is live. corexi.com/mcp.

See you next week.

Let me say this openly: for a while now, every time I open LinkedIn, I feel a kind of fomo (fear of missing out)

I see too many launches. I read posts from people who’ve never even built a website in their lives saying “I built this product in this space, that product in that space.” On one hand it excites me, makes me happy, makes me feel like I’m witnessing history. I’d say the AI revolution is the third major step after electricity and the internet. But at the same time something starts whipping me from the inside: “I should ship something under my venture studio Keiki, fast. I need to catch up.” And it whips hard.

This piece is partly an attempt to make sense of that urge. But first, let me tell you something that happened two days ago.

I came to Ordu from Barcelona to see my family. Ordu is a small but fast-growing town on the Black Sea coast in northern Türkiye, sweet and still keeping its niche character. Look it up if you haven’t.

The night I arrived, I sat down at one of my favorite pubs, Köşk Pub. I was going to have a beer and some fried peas. When I scanned the QR code to order, there it was: an online menu made with AI. I noticed it the moment I saw the UI. Digging in a bit more, I found bugs, structures that didn’t quite fit, and a section that redirected me to Replit. It was built with Replit.

Picture this. In Ordu, in a local pub, there’s a product they built themselves with Replit. In the 50-year history of Köşk Pub, perhaps the biggest technological transformation, made possible by AI.

I got up immediately and walked the three steps over to the owner. I asked him: “That’s interesting. Why did you change the old online menu and build this one with Replit?”

He used to pay 100 euros a year to the previous provider. He’d heard about Replit, tried it, built it easily. As he was telling me this, his face was smiling, he was visibly proud of himself, and I could tell he was glad I’d noticed.

“So how much did you spend?” I asked.

“150 euros,” he said.

Hold on a second. He could have kept paying 100 euros a year for a ready-made online menu provider with all the content already set up. Instead, he sat down, spent his hours, went through the learning curve, paid more money, and built it himself. And he’s happy about it. He’s proud of it.

Does this remind you of something? It reminded me of two things directly: cake mixes and IKEA.

People don’t buy ready-made cakes from the supermarket. They buy cake mixes where you crack the egg, stir the ingredients, bake it yourself. They pay more for it. And they enjoy doing it. Because we’re human, craft and the “I made this” feeling still matter a lot.

Replit isn’t just selling a browser-based, AI-powered product development platform. It’s giving you the ability to craft. It’s pulling you into the work. Especially for people who’ve never built products before, people who don’t see themselves as “product people,” this is incredibly valuable.

Today, Replit has become the cake mix of digital products.

When I noticed the Köşk Pub story, I realized this is actually the concrete, street-level version of the launch wave I see on LinkedIn.

On one side, a pub owner in Ordu. On the other, people all over the world shipping launches every week with Cursor, Replit, Lovable, Bolt, v0. Both move with the same urge. You don’t subscribe to Notion anymore. You build your own Notion, the way you like it, the way it fits you. Yes, you spend more than what you’d pay Notion over years, but it becomes your Notion.

I want to call this self-product. Not because it’s micro. It can be small or big. The intent is what’s different. Not built for the average needs of a generic crowd, but for your own problem.

The self-product era has started.

So, do you think this just keeps going and we end up with quadrillions of apps, digital products, everyone using their own branded product for everything? Doesn’t this make things a bit chaotic? Wouldn’t it have a destructive effect on some industries? Are everyone’s products actually high quality, working properly, a sensible investment?

Put another way: if everyone builds a solution to their own problem, will all problems be solved?

The answer is no.

Some will, yes. Köşk Pub’s menu got solved. That guy’s 100-euro SaaS dependency got solved. I’m not dismissing that, I mean it. But two new problems emerge at the same time.

The first is staying invisible. Building used to be hard, getting noticed was relatively easy. What you put on Product Hunt was actually competing against a small number of alternatives. Now it’s the other way around. The launches we see on LinkedIn are the ones that managed to be visible. But behind every visible launch, there are tens, hundreds of half-finished, abandoned, unused products. We don’t see them because they didn’t get visible. Classic survivorship bias. When you see “I shipped last week” posts in your feed, what you need to keep in mind is: there were hundreds of other launches you didn’t see.

The ability to ship has become baseline. “I shipped” doesn’t mean anything on its own anymore. The new bottleneck is this: can you get people to see what you built, try it, and stay with it? Distribution, attention, mindshare. These used to come after the product. Now they weigh as much as the product itself.

The second is the junk sea. App Store and Product Hunt are already full. They’ll be 10 times fuller in the next 5 years. Simple arithmetic. As the marginal cost of building approaches zero, production becomes unlimited. Anything that becomes unlimited loses value. We’ve already lived through content inflation, and we’re still living it. Product inflation is coming too. Some of it will be personalized, meaningful self-products. But most of it will be junk. Half-finished, buggy, abandoned, security-flawed products that nobody maintains. This junk might solve someone’s problem for a moment. But it doesn’t create lasting value. It floats around like a ghost.

There’s also this: does Köşk Pub’s menu actually belong to him, or to Replit’s template? We changed the tool, but the dependency just moved. What will the pub owner do if a bug shows up in the menu six months from now? Building is one thing. Sustaining is another.

The notion of “one right product” has collapsed. I accept this, and I don’t think there’s any going back. Notion doesn’t have to be the same for everyone anymore. This means more variety, more choice, more personalization. Good.

But what’s replacing it, the self-product, isn’t enough on its own. New filters emerge: Is it visible? Is it sustainable? Is it memorable? Can it actually be owned?

A product is no longer just an artifact. It’s something that forms around voice, attitude, community, and story. As building gets easier, everything outside of building gains weight.

I don’t know what state Köşk Pub’s menu will be in two years from now. But I do have a prediction about where digital products are headed in two years. Bookmark this if you want, let’s see in 2028.

Here’s what will win: products that can actually personalize. I’m not talking about changing the color of the interface. I’m not talking about the surface-level “personalization” we’ve been discussing for the past decade. Products that understand you, shape themselves to you, in another dimension. Products that carry AI not as a feature, but in their veins, their voice, their power. Products that start to resemble you the more you use them.

I’m not sure if fomo is needed. Even though it feels like we’re horses being whipped on our backs in a constant time-race, most tokens will end up as junk. As always, what should be will be. The right things will win.

With one difference: the problem of this era is no longer just “doing the right thing.” What can be distributed will win. As what can be built gets built by everyone, what can be noticed will stay scarce. Being good is the baseline. Being reachable is the new moat.

Köşk Pub’s owner built his own menu for 150 euros, and he’s proud. In the next 10 years, millions of people around the world will do something similar.

Most of these will be forgotten, abandoned, decayed within two years. Some will last. Some will become important parts of people’s lives. We don’t know yet which will be which, but where it lands depends on whether people can convert the “pride of building” into the “investment of sustaining.”

I’m writing this while still in Ordu. Every time I open LinkedIn, I’ll feel fomo again. But a different fomo. No longer “I should ship something too.” Now: “will what I ship be visible, sustainable, meaningful?”

The self-product era has started. The story isn’t written yet. We’re writing it together.

Last week I told you we had given birth to the product. This week the product got its name. Not the brand name. Corexi has been Corexi for a while. The sentence that explains what it actually is.

Corexi is a continuous UX layer for AI-native products. It analyzes them, keeps analyzing them, and keeps improving them.

Thanks for reading keiki weekly! Subscribe for free to receive new posts and support my work.

That sentence took longer to land than the product itself. Once it did, everything around it snapped into place: the roadmap, the vision, the homepage, the way I talk about it in calls. Corexi is no longer a tool with a pitch. It’s a category. A UX category for the AI world that didn’t exist before, at least not as a continuous layer running on top of AI-native products.

Updated vision and roadmap: https://www.corexi.ai/en/vision

+ studio updates

1. Roadmap, finalized. The keiki portfolio is now locked at 12 products. Sequenced, not simultaneous. Corexi first, then the rest, one at a time.

2. keiki.studio, quieter on purpose. We had a “product interaction zone” on the site. A build-in-public space where visitors could follow each product’s roadmap step by step and vote on what to add next. It’s hidden for now. We’re early enough in the story that too many moving parts read as noise, not transparency. The goal this week was a cleaner first impression: who we are, what we’re doing, why it matters. The interaction zone will come back, but later, when the story can carry it.

3. Trademark update. Quick recap for new readers: a Cyprus-based company called Benipol, which only makes baby products, had registered “keiki” as a trademark. When we moved to use “keiki studio” in our own category, they objected. We’ve now formally assured the EUIPO that we have no intention of operating in baby or child products, in any form. Now we wait.

+ venture lab (product) updates

This week was Corexi, and only Corexi. As the solo founder on the venture side, I made a deliberate call: until Corexi matures a little, no new products go live. The blurple studio team is behind me, but the venture side is mine to carry, and carrying more than one product at a time is how good products get half-built.

1. Corexi got its sentence. As above. The shift from “a product I’m building” to “a continuous UX layer for AI-native products” rewires everything downstream. Roadmap, vision, pricing logic, partner conversations. All cleaner now.

2. Vision and roadmap, locked. Corexi is a continuous layer, not a one-off audit. Continuous analysis. Continuous improvement. That’s the whole shape of it.

3. Website updates, in plain language.

   • The site now reads, top to bottom, as a continuous UX layer for AI-native products. The whole homepage has been rewritten around this positioning.

   • Added the “two moments” message and polished the homepage flow. Cleaned up a leftover conflict between an old “Same engine” pill and the new “Two Moments” section.

   • Vision and story pages rewritten to match the new positioning.

4. Demos and partners. Demo slots for the week are fully booked. We’ve also started onboarding the first partners. The early signal is good. People understand what Corexi is faster now that I can say it in one sentence.

A Note on Focus

This week tested the discipline of doing one thing at a time. The temptation to start a second product is always there, especially with a roadmap of twelve sitting in front of you. But solo founder math is unforgiving. Every new product is a tax on the one that’s almost ready. Corexi gets the gas until it can stand on its own. Then the next one.

See you next week.

Thanks for reading keiki weekly! Subscribe for free to receive new posts and support my work.

Welcome to the twelfth edition of keiki weekly. This week is different. This week, we launched.

+ product updates

1) Corexi: Live and in Closed Beta

Two weeks ago, I wrote about going from zero to a working product in one week. This week, that product went live.

Thanks for reading keiki weekly! Subscribe for free to receive new posts and support my work.

Corexi is now in closed beta. Real users. Real products. Real feedback coming in.

Let me tell you what these two weeks actually looked like.

https://www.corexi.ai/en

Week one was pure build. Architecture, core scanning engine, PX Score methodology, dashboard, findings, fix code output. The kind of week where you wake up thinking about component detection logic and fall asleep thinking about scoring weights.

Week two was refinement, launch preparation, and the hardest part: deciding what’s good enough to ship. Landing page, product positioning, beta banner, responsive fixes, PH page, LinkedIn content, video recording. The unglamorous work that separates products that launch from products that stay in Notion forever.

What is Corexi?

Corexi is an AI-native product experience intelligence platform. Not a UX audit tool. A continuous intelligence layer that lives with your product, learns it, and fixes it.

Three layers work simultaneously:

Visual AI: Every page scanned at desktop and mobile viewports across 9 UX categories. Navigation, visual hierarchy, CTA effectiveness, form usability, content readability, color and contrast, spacing, mobile compatibility, interaction design.

Behavioral data: Connects to your existing tools, GA4, Hotjar, Clarity, Amplitude, Firebase, and reads real user behavior anonymously. No new tracking code. No PII. GDPR compliant. The more sources connected, the sharper the diagnosis.

Corexi AI Engine: Fuses both layers using 200+ peer-reviewed UX studies. Nielsen heuristics, Gestalt principles, WCAG standards, cognitive psychology research. Not a black box, every score is traceable to its research source.

Industry first: ADHD and neurodiversity compatibility scoring. A page can pass every accessibility check and still overwhelm a neurodivergent user. Corexi measures this automatically. No other tool does.

The output: a PX Score (0-100) broken into five dimensions, visual quality, behavioral health, usability, accessibility, cognitive load. And for every finding, fix-ready code or an IDE-specific prompt for Cursor, Claude Code, or VS Code.

Not a report. Action.

Early Users

We didn’t launch cold. Keiki Studio is built alongside Blurple Studio, our sister AI-native product studio that has spent years designing and building digital products for companies like ZES, Electrip, and several European brands, touching the experiences of millions of users. That deep domain expertise is baked into Corexi’s methodology.

Our first closed beta users are from that same client base. Companies who already trust us with their products, who give honest feedback and push back when something doesn’t work. That’s exactly the environment we needed to stress-test Corexi in the real world.

What’s Live Now

Full site crawl, visual AI analysis, behavioral data fusion, PX Score with 5 dimensions, finding cards with severity ranking, fix-ready code output, IDE selector (Cursor, Claude Code, VS Code, Windsurf, Replit), neurodiversity lens, PX Badge eligibility, continuous scanning with frequency settings.

What’s Coming

Competitive Benchmark, Predictions, PX Product Directory, and the long-term vision: Autonomous PX Engine, a closed-loop system that detects, generates fix variants, tests with synthetic users, and deploys improvements. Continuously.

2) Pawspan: Next on the Runway

Corexi is live. Now it’s time to turn our attention to what’s next.

Pawspan, our AI-native pet longevity and wellness platform, is moving from parked to active. With Corexi now in users’ hands, we’re ready to start building again. The co-founder conversation is progressing well and the energy around the product is strong. Expect proper Pawspan updates starting next week.

A Note on Shipping

Launching is humbling. You spend weeks building something, and then you put it in front of the world and realize how much you still don’t know. Which message lands. Which feature gets ignored. Which assumption was wrong.

That’s the point. You can’t learn any of that in Notion.

If you build digital products and want to understand your users better, not just what happened, but why, and how to fix it, I’d love to have you in the beta.

Completely free. No credit card.

👉 corexi.ai/en

See you next week, with real data and real learnings.

Thanks for reading keiki weekly! Subscribe for free to receive new posts and support my work.

Welcome to the eleventh edition of keiki weekly. This is our second Monday edition, and it already feels like the right rhythm. This past week was one of the most intense we’ve had: pure product development and deep focus. No distractions, no context switching. Just building.

Thanks for reading keiki weekly! Subscribe for free to receive new posts and support my work.

Thanks for reading keiki weekly! Subscribe for free to receive new posts and support my work.

+ studio updates

1) Toward a Unified Ecosystem

We’re exploring bringing blurple studio and keiki studio together under a single blurple ecosystem. Two studios, one shared foundation: client work and our own products living under the same roof. Still early thinking, but the direction feels right. More on this as it takes shape.

+ product updates

1) Corexi: From Idea to Live Product in One Week

This was the week. I built Corexi from zero to a working product in seven days.

Now, I need to be honest: saying “one week” doesn’t tell the full story. The idea, the architecture, the scoring methodology, the positioning have been living in my head for over a year. What happened this week was the culmination of that thinking, combined with months of building experience with AI-native tools. The coding took one week. The product thinking took much longer.

What is Corexi?

Corexi is an AI-Native Product Experience Intelligence platform. It opens a new lens into your users’ unknown experience by combining three layers:

• Visual layer (Capture): Multi-model vision analysis crawls your screens across desktop and mobile viewports. Connected analytics (GA4, Mixpanel, Clarity, Amplitude) feed behavioral data. All signals land in one place.

• Data layer (Diagnose): A proprietary hybrid reasoning engine fuses behavioral data with visual findings, scores 9 UX categories, and maps friction to real user impact. This is the Corexi AI engine, UX-trained and research-backed.

• Fix layer: Severity-ranked findings with annotated evidence screenshots and stack-aware copy-paste code. Ship the fix, not the ticket.

The key differentiator: Corexi doesn’t just show you what’s broken. It writes the fix. Every finding comes with implementation-ready output: direct code or IDE-compatible prompts for Cursor, Claude Code, Windsurf, Replit, GitHub Copilot, or VS Code.

From URL to scored, fix-ready report. Under 30 seconds. No SDK, no tag manager.

PX Score: Your Product’s Experience Quality, Scored

Every scan produces a PX Score (0-100), broken down into Visual Quality, Behavioral Performance, Usability, Accessibility, and Cognitive Load. This isn’t a vanity metric. It’s a benchmark you can track week over week, compare against competitors (coming soon), and use to communicate UX health to stakeholders in a language they understand.

PX Badge: Verified Experience Quality

Products that maintain stable 80+ scores across at least 4 scans with connected analytics earn a verified PX Badge. Think of it as a trust seal for experience quality. You can embed it on your site or pitch deck. It’s not awarded by a jury. It’s earned by your data.

Intelligence That Deepens With Every Scan

Corexi doesn’t just audit once. It learns your product. Each scan deepens a behavioral profile. It tracks what changes. It predicts what’s next. The more data flows in, the sharper the diagnosis becomes. 9 research-backed categories, continuous understanding.

What’s Coming Next

We shipped the core product, but the roadmap is already taking shape:

• Competitive Benchmark (Coming Soon): Compare your PX Score to the peer median for SaaS in your region. See where you lead or lag.

• Predictions (Phase 2): Industry-aware trend forecasting. Seasonal spikes, emerging UX patterns, and proactive recommendations before issues hit your metrics.

• Test + Compare (Phase 2): Generate fix variants and validate them with synthetic AI users. Measurable outcomes, not opinions.

• Autofix Agent (Phase 3): One-click fix execution via Cursor, Claude Code, Replit or terminal. Apply a suggested patch into your repo in one action. No copy-paste loop.

2) Pawspan: Parked, Not Abandoned

We paused Pawspan development this week to go all-in on Corexi. Partnership conversations with our potential co-founder are progressing well and the energy remains strong. But shipping one product right matters more than splitting focus across two. Pawspan returns once Corexi is live.

3) Launch Week Is Here

Corexi is targeted to go live this week. We’ll announce on Product Hunt and across our social channels. Stay tuned.

A Note on Focus

One thing this week reinforced: the sequential approach works. When you try to build two products at once, you build neither well. Parking Pawspan wasn’t easy, but Corexi needed full attention to reach the quality bar we set for it. One product at a time. Ship it right. Then move to the next.

See you next week, hopefully with a launched product and real users.

Best,

Thanks for reading keiki weekly! Subscribe for free to receive new posts and support my work.

Thanks for reading keiki weekly! Subscribe for free to receive new posts and support my work.

Welcome to the tenth edition of keiki weekly. Yes, tenth. Yes, it’s been two months. No, I wasn’t on vacation.

The short version: I killed two products, had conversations with investors including Google Ventures, restructured the entire studio, and came out the other side with sharper focus and two new products that actually make sense.

Thanks for reading keiki weekly! Subscribe for free to receive new posts and support my work.

The long version is below.

Thanks for reading keiki weekly! Subscribe for free to receive new posts and support my work.

+ what happened

Let me be honest. The last two months were messy in the best possible way.

When you’re a solo founder running a venture studio, you don’t have the luxury of a board meeting to validate your direction. You have your gut, your data, and the occasional conversation that changes everything.

Here’s what changed:

I killed our UX copilot product. The tool we’d been building to help teams fix UX issues faster. It worked, but it was playing in a crowded copilot space with thin differentiation. The more I used it, the more I realized the real problem isn’t helping people fix UX issues. It’s helping them find and measure them continuously. Different problem, different product. So I killed it and started over.

I killed Houra AI. Our meeting scheduling product. I had serious conversations with major VCs, including Google Ventures. The feedback was consistent: the timing isn’t right, the market is brutally competitive, and the moat is thin. When multiple smart investors tell you the same thing, you listen. I unpaused the idea. Maybe it comes back someday. Not now.

I restructured the studio. Keiki is now leaner, faster, and more focused. Some decisions were hard. Some were overdue. The result: clarity.

Every venture studio goes through this. Especially when there’s a solo founder making the calls. Paths shift, products die, new ones emerge. That’s not failure. That’s the process.

+ what I learned

Two months of killing products, talking to investors, and rebuilding from scratch taught me a few things worth sharing:

• “Copilot” is a crowded word. Every AI product calls itself a copilot now. The moment your positioning sounds like everyone else’s, you’ve already lost. I moved from “helping you fix” to “measuring and detecting for you.” Different verb, different category.

• VCs don’t kill your idea. They sharpen it. The Google Ventures conversation didn’t discourage me. It forced me to ask harder questions about moat, timing, and market readiness. Every “no” made the next product better.

• Subtraction is a feature. I went from 10 ideas in the pipeline to 2. The portfolio looked impressive on paper. In practice, it was a distraction factory. Two focused products with real development time beats ten concepts with fancy landing pages.

• Solo founder ≠ alone. I still run blurple studio with an incredible team. The venture side is solo, but the thinking isn’t. Every product decision gets pressure-tested against real client projects and real user pain.

+ what’s next

Two new products. Both in stealth for one more week.

Product 1: UX, but intelligence this time.

I took everything I learned from our UX copilot and from 20+ projects at blurple studio, and I’m building a product experience intelligence platform. Not a copilot. Not an audit tool. A continuous intelligence system that measures, scores, and tells you exactly what’s wrong with your product’s experience, with the code to fix it.

Same problem space, completely different category. Landing page goes live next week.

Product 2: AI meets the pet industry.

This one might surprise you. I’ve been quietly researching the pet health and longevity space for months. There’s a massive gap between what pet owners need and what technology offers them. I’m building an AI-native product to close that gap.

Landing page also goes live next week.

Two products, two very different markets, same AI-native DNA.

+ studio updates

blurple studio is running strong. Multiple active projects across mobility, energy, and fintech. The team is solid. We recently completed a full repositioning (v5.5) as an AI-Native Product Studio, sharpened our service lines, and updated everything from pricing to proposals. The studio side funds the venture side. That hasn’t changed.

keiki studio is now fully focused on two products instead of a scattered portfolio. Real development hours, real progress, real deadlines. Not theoretical roadmaps.

+ the return of this newsletter

keiki weekly is back. Every Monday.

The difference from before: more signal, less noise. Two focused products instead of five scattered ideas. Real progress updates instead of vision documents.

If you’ve been here since issue #1, thank you for sticking around through the silence. If you’re new, welcome to the build.

I’m curious: what would you rather see in these updates? More technical build details? More business/strategy thinking? More honest founder reflections? Hit reply, I read everything.

See you next Monday with two landing pages and a lot more to talk about.

Best!

Thanks for reading keiki weekly! Subscribe for free to receive new posts and support my work.

Replit was born as a “vibe code” platform. But today with Agent 4, we see it transforming into a comprehensive digital production tool, or rather, a fully-fledged product creation machine. I think in the coming months, this structure will cover the entire process of a digital product from its birth to its continuation. I mean from research to design, from development to growth, and then analyzing itself internally to become a comprehensive product body that heals itself.

I’ve been using Replit for over a year and I’m a firsthand witness to its evolution. I’m actually a user who frequently gives feedback and constantly bugs the platform. For the last 5 months, I’ve moved all my digital assets, my studios’ websites, and my studio’s AI-powered operating system (blurple OS) completely to Replit or built them directly on it. So Replit has already positioned itself for me as a hub where I combine, sustain, and continue to develop all my digital products.

I want to clarify something here. For digital projects that will truly become independent, pivot seriously, and grow on a heavy infrastructure architecture, I sometimes start them in Replit just to do a POC (Proof of Concept) and then move them to the Cursor environment. Or I just start these types of projects directly in Cursor. That’s the reality of the job.

But from getting quick outputs to maintaining my continuous platforms and micro-products I mentioned above, I keep them all in Replit. We’ve even made this a policy for our company, you could say. Because why wouldn’t we? Is there a more reliable development environment or infrastructure that better matches our needs right now? (There is still no European location for data center usage... if they add that too, it will be truly flawless.)

The main thing I want to talk about in this article is Replit Agent 4. We really need to think and write comprehensively about this update. Within the AI development world (call it what you want, I don’t really like saying “vibe code” because I think it makes the topic too light), we clearly see that Replit continues to carry the pioneering torch. The reason I kept this introduction so long was to point out how much importance Replit carries in the lives of me and people like me who develop products and have figured out the true power of AI.

Now with Agent 4, Replit has evolved into something we can’t quite describe yet. Because some guys in California or Tokyo haven’t invented a cool word for it yet. For now, I’ll call Replit a “product development and survival ecosystem”. We could even call it the new operating system of our lives for people like us.

I might sound like I’m talking too ambitiously. But what Agent 4 offers proves how much more Replit will position itself at the center of our digital lives and go even deeper. The management team handled this transformation with such calm, professional, and cool steps that it feels to me like they will go beyond the AI world and become our new Apple. I haven’t praised a brand like this in a long time, believe me, I’m surprised too.

So what will we be able to do with Agent 4? Or rather, how will we redefine that painful process we call “product development”? Let’s look at that.

The Collapse of the Wall Between Design and Code

I was thinking about this while walking Liza the other day. For years, there has been a meaningless translation problem between designers and developers. We get an idea, we draw a great interface, but that design loses its soul when it’s poured into code. Or that process drags on and on. With Agent 4 and what they call “Design Freely”, they’ve moved to a structure that does visual production and code production at the same time. Imagine an infinite canvas. While you play with the visual details of the app on one side, the Agent continues to build the code in the background while you refill your coffee. Interface and infrastructure are no longer two different worlds. You can instantly integrate that brilliant design idea that came to you in the shower into the product. The hassle of drawing first, getting approval, then handing it to the developer... meaning that terrible friction is finally over.

Escaping the Heavy Lifting and Focusing on “Intent”

Setting up user authentication systems, connecting databases, making the backend talk to the frontend... These used to be mechanical tasks that turned into late-night note-taking struggles, eating away at our brains. Agent 4 clearly says, you just put your intent out there, and I will handle the database, the background, and the frontend simultaneously with parallel agents. The point here isn’t just speeding up the work. As a team, we stop thinking “how do we code this”. Instead, we put our minds to “what will the user feel on this screen”. While Agent 4 does the heavy lifting of code and coordination, it leaves us only the creative strategy that comes with being human.

Not Just Software, A Comprehensive Product Ecosystem

This part is critical. We aren’t just talking about pushing out a web app. Mobile apps, landing pages, pitch decks, and even videos... It can produce all of them within the context of the same project. Marshall McLuhan says “The medium is the message”. This medium of Replit gives us this message: The product is no longer a combination of code blocks written line by line, but a single reflection of your intent and vision in digital form.

The Ultimate Proof: Bringing the Giants Under One Roof

If you want concrete proof that Replit is evolving into this comprehensive product life cycle machine, just look under the hood at their subprocessors list. I was checking it recently and it is literally a who’s who of the tech world. Anthropic, OpenAI, Stripe, Twilio, Cloudflare, Datadog...

Replit has brought all these massive, independent players together. You no longer need to run around opening a dozen different tabs or trying to stitch APIs together like a mad scientist. They have gathered the biggest infrastructure and service giants and made them applicable from a single, unified point. This is exactly what proves they are building a true living hub, not just a coding interface.

So Where Are We Going From Here?

Remember at the beginning of the article I mentioned a comprehensive product body that will heal itself. Agent 4 is exactly building the nervous system of this organic structure. Today it builds code and design at the same time. Do you know what will happen tomorrow? After the product goes live, it will monitor user behavior. It will say “Users can’t find this button, there is friction here”, and without you even noticing, it will redesign that invisible interface and push it live instantly. It will run its own tests, it will pour its own growth strategy into code itself.

Instead of mimicking the clumsy software processes of the old world with AI, they are building a completely intent-based, organic, and breathing product infrastructure. In short, we aren’t just building things that run on a screen, we are rewriting the genetics of our digital assets. And yes, if they also solve that European data center issue, we will have a truly flawless digital home.

Welcome to the ninth edition of keiki weekly. This week has been characterized by deep focus and internal refinement. While there were no loud external announcements, the engine room has been busier than ever. We stepped into a “feedback week,” turning our lenses inward to polish our flagship product before its imminent public debut.

We are currently in the final stretch the calm before the launch, solidifying our product’s logic, user experience, and strategic positioning.

Thanks for reading keiki weekly! Subscribe for free to receive new posts and support my work.

+ studio updates

1) The Digital Home: Agentic Focus & Gratitude

Recently, we overhauled our web presence with a distinct focus on an “Agentic” approach, detailing our AI solution areas and services. The response has been overwhelmingly positive.

• Gratitude: A sincere thank you to everyone who visited and provided feedback. Your insights validate our direction.

• Invitation: If you haven’t explored the new site yet, we highly recommend taking a look to see how we’ve structured our AI-driven service architecture. Here: https://www.keiki.studio/

Aside from these digital housekeeping items, the studio side remains stable as we funnel our energy into the Venture Lab.

+ venture lab {product} updates

1) The Flagship (Formerly Fluxar): Preparing for the “Sense” Era

We have officially updated the name of our flagship product (Fluxar). However, we are holding back the reveal for just a little longer. Our target is to share the new identity simultaneously with the v2.0 launch in approximately 10 days.

This week was dedicated to rigorous internal feedback loops. Together with my strategic partner Selen, we dissected the product’s Information Architecture (IA) and User Experience (UX).

• Micro-Refinements: We polished the landing page content and overhauled the user panel’s navigation. Specifically, we refined the Global and Local menus to ensure a seamless, intuitive flow for the user.

• The “Sense” Integration: v2.0 will introduce the “Sense” capability. This allows the product to ingest analytics data from mobile, web, desktop, and even kiosk applications.

• The Power of Context: By combining our existing AI Engine (already strong in UX analysis) with real-time analytics data, v2.0 will be truly data-driven. It will recognize the user, understand the product’s context, prioritize actions, and provide comprehensive, change-sensitive insights.

• Human-Expert Loop: We are bridging the gap between AI and human intuition. Users can now opt for feedback from a real human UX expert—either as a one-off service or a recurring subscription via the panel.

2) Technical Health & Revenue Strategy

While we refine the frontend, the backend is undergoing heavy stress testing to ensure stability at scale. Simultaneously, on the business side, Selen has taken the lead on defining the revenue model, packaging strategy, and sales approach. We are in the final evaluation phase before locking in our Go-To-Market (GTM) strategy.

3) On the Horizon: Houra AI & New Concepts

• Houra AI: Currently, this product remains on the shelf, but it is constantly in our peripheral vision. The plan is to unpause development and fast-track it immediately after the flagship v2.0 launch (target: 10 days).

• Incubating a New Concept: We have started the documentation phase for a brand-new product. It sits on a similar axis to our flagship—focused on user feedback for digital product companies. We are currently defining its scope and strategy and will share details once the blueprint is finalized.

A Note on The Podcast

Last week, we introduced the audio companion to this newsletter, and we are continuing that tradition. While this text provides the structured report, the podcast allows us to convey the emotion, the “story,” and the nuances of our journey.

If you want to hear the voice behind the vision, you can listen to this week’s episode via the link below.

See you next week hopefully with a new name and a launched product.

Best!

Thanks for reading keiki weekly! Subscribe for free to receive new posts and support my work.

Welcome to the eighth edition of keiki weekly. This week, while we polished the face we show to the world, we focused heavily on solidifying the studio’s “invisible” layers: our legal and operational foundations. Pausing to check our structural integrity while moving at speed is the most critical part of our long-term vision.

We are also on the eve of new strategic partnerships regarding our product development philosophy and an exciting “rebirth” process for our flagship product.

+ studio updates

1) Structural Clarity and Digital Experience

This week, we completely clarified keiki studio’s legal and internal operational processes. We are now moving forward on much firmer ground. Following the major updates to our website in recent weeks, we made User Experience (UX) our focal point this time.

• Split Interface: We added two main sections to our homepage that allow visitors to distinguish between our two main divisions (Venture Lab & AI Solutions) at a glance and navigate accordingly.

• Micro-Improvements: We completed micro-updates to boost web performance and smooth out the user flow.

2) Replit Strategic Partnership and Development Philosophy

Taking our long-standing warm relationship with Replit a step further, we started a special partnership dialogue this week. We will share details in the coming weeks, but I would like to touch upon our development philosophy as an AI-Native studio:

• Cursor vs. Replit: While we actively use Cursor for our products requiring complex, deep engineering, our history with Replit goes back to the platform’s early days.

• The Journey from 0 to 1: We believe in Replit’s secure and agile structure for projects that require speed, are in the validation phase, or are moving from 0 to 1. Replit is unrivaled when it comes to “safely building” a project.

• Hybrid Approach: We follow hybrid paths depending on the project; we often start fast and reliably with Replit, retaining the flexibility to migrate to different tools as the project matures.

+ venture lab {product} updates

1) v2.0 and the “Sense” Phase

We have gained serious momentum on the product side. I can say that we have completed the v2.0 leg. Currently, our system is in the “Sense” phase and can connect with other analytics tools to offer the user beneficial and actionable insights.

• Human-Expert Loop: Development continues to integrate not just AI, but also human expertise (the human-expert loop) into this game.

2) Name Change: A New Identity is Born

Sometimes legal necessities birth creative opportunities. Due to trademark issues, we decided to change the name of Fluxar. The team has been heads-down on this for the last three weeks, and we have finally selected the “right” name that fully reflects our vision and truly resonates with us.

• Launch: When we share v2.0, we will do so under this powerful new identity.

3) Revenue Model and “Orbit” Strategy

We are clarifying the strategic planning for the launch of “Orbit,” which will present the product’s core value proposition:

• Validation: Leveraging the experience of our other company, Blurple Studio (a seasoned UX Studio), we have already started our initial tests with existing clients.

• GTM (Go-to-Market): We are finalizing how the first public launch will be structured regarding subscription tiers and free trial models, and which brands we will collaborate with initially.

See you next week with our new name and finalized partnerships.

Best!

Welcome to a special combined edition of keiki weekly. As we complete the first weeks of 2026, we wish everyone a year beyond their dreams. We decided to merge our sixth and seventh weeks to balance a short holiday break with intensive planning and infrastructure refinement. This allowed us to focus on long-term stability rather than just weekly output.

+ studio updates

1) Technical Infrastructure and Website Optimization

We conducted a comprehensive analysis of the keiki web experience. Our focus was on the “invisible” layer that dictates performance.

• Performance Overhaul: We implemented updates ranging from server selection and CDN consistency to metadata settings and accessibility.

• Backend Stability: These technical refinements ensure a smoother and faster experience for every visitor across all devices.

2) Agentic keiki: Enhanced Interactions

Our website agent has been upgraded with a more robust knowledge base and a clearer mission to guide our visitors.

• Guided Question Sets: We now offer ready-made question sets based on our refined services, helping users find what they need instantly.

• Smart Navigation: The Agent now directs visitors to relevant pages within our site using direct links based on their specific queries.

• Lead Capture: If the Agent determines a human touch is needed, it now requests contact information to bridge the gap between AI interaction and a keiki team member.

3) New Year Eve: The 25-Minute Project

On the morning of December 31, 2025, we launched a micro-tool as a celebration of AI’s agility.

• AI Oracle: Developed in just 25 minutes, the New Year Eve tool serves as an AI oracle, providing hope, positivity, and industry data for 2026.

• Action-Oriented: This project reflects our philosophy that if an idea is perfectly timed for the current AI era, it should be shared immediately rather than waiting for another year.

• Market Insights: The tool even projects industry growth, such as the micro AI tools market reaching $10 billion by 2026. You can still try it at keiki.studio/newyeareve.

  

4) AI Solutions Zone: Service Frameworks

We have clarified our AI-Native and AI-First service frameworks. We now openly share our delivery timelines for the AI era to ensure full transparency with our partners. You can explore the updated zone at keiki.studio/ai-solutions-zone.

+ venture lab {product} updates

Fluxar: Identity, Integration, and Professional Workflow

The last two weeks were dedicated almost entirely to Fluxar, focusing on a clearer identity and deeper analytics integration.

1) New Experience and UI

• Landing Page: We designed a brand new landing page to communicate our value proposition more effectively.

• One-Shot Reports: The report experience has been redesigned from the ground up to offer a cleaner UI and improved information hierarchy.

2) Analytics Platform Integrations

• Microsoft Clarity: Full integration is complete. Fluxar now analyzes heatmaps, session recordings, and user insights to produce UX recommendations.

• Multi-Platform Intelligence: Analytics sources are now dynamically filtered by project type. Web projects see GA4 and Clarity, while Mobile and Kiosk projects see platforms like Firebase and Mixpanel.

• New Platform Support: We have added Segment and PostHog to our UI, with API integrations currently in the development phase.

3) Dashboard and UX Refinements

• Report Management: The latest report is now featured at the top with precise timestamps, while previous reports are gathered in a collapsible section.

• Standardized Data: All metrics and scores now follow a consistent format with one decimal point to improve readability across all comparisons.

4) Development Workflow (Railway Integration)

• CI/CD Pipeline: We have transitioned to Railway for production deployment, enabling an automated workflow through GitHub integration.

• Environment Management: Development, staging, and production environments are now clearly separated with centralized environment variable management.

The momentum is building as we approach our v2.0 Public launch. In our upcoming eighth week, we plan to activate the Human-In-The-Loop system for final controls. My strategic partner with and I as we prioritize the roadmap and finalize the business model for the next phase.

See you next week!

Thanks for reading keiki weekly! Subscribe for free to receive new posts and support my work.

I have been following Andrej Karpathy (TIME100 Most Influential People in AI) with great interest for some time. His philosophical and futuristic approaches to AI, combined with hearing his name frequently from various circles lately, placed him right at the center of my focus. (I generally believe that there is no one we cannot reach) Recently, while returning from a New Year’s holiday trip from Rome to Barcelona, I had the chance to listen to plenty of podcasts during that 1000 kilometer journey. That was when I came across the High Agency episode of ’s 5 Kişi Podcast.

What is High Agency? In its simplest form, it is the ability to find the inner drive to go around, under, or through a wall when you encounter one, instead of just saying the road is closed. According to Andrej, this is the most critical and defining skill an individual can possess in the AI era. Since technology now provides everyone with the same baseline of power, the real differentiator becomes your agency, the force that decides where to direct that power.

They say a person is always searching for themselves. For me, life is a quest and a journey; I am always searching for myself and I continue to do so. Every now and then, when I find something about myself, I stop and say “Eureka!” to rest for a while. I reflect on what I found, shoulder what I have learned and clarified, and set out on the road again. The concept of High Agency became one of the things I fitted to myself during this journey of searching. I can say it feels much better than fitting myself to ADHD, at least :)

Without straying too far from the subject, I want to share a thought I expressed recently that my girlfriend liked very much: “I try to fix certain things in my life to ensure the things that need to move, actually move.” Everyone can derive their own meaning from this sentence; I am curious to hear what you understand from it, so feel free to write to me.

Anyway, while on the road with these thoughts, I tried many different combinations and finally reached Andrej Karpathy! Being careful not to overwhelm him, I asked him the critical questions that were on my mind.

A Lever or a Crutch?

I asked him whether the concept of High Agency is a lever for humanity and the individual, or a crutch. His answer was very clear:

“AI is a superpower for high agency individuals. It allows what used to take a person weeks to do to be finished in an hour now. However, for low agency individuals, AI carries the risk of turning into a comfort zone that removes the necessity of thinking. The difference is determined by the individual’s ‘hunger for problem solving, curiosity, and the strength to take initiative’ rather than the technology. In other words, AI does not atrophy our muscles; on the contrary, it frees them to focus on larger and more complex problems.”

This response shows that the concept of High Agency is not limited to entrepreneurship; it also opens a new door of opportunity for us. I believe the way to be a high agency individual is to ensure that one does not surrender to inertia and is able to maintain that challenging focus.

The Vertical Future and the Soul of the Product

My other question was about the LLM OS vision, where general models become a standard commodity, and where the true soul of a product would reside in such a world. Andrej said:

“Models will become like electricity over time; they will be everywhere and will work with the same quality for everyone. In this world, the soul of a product will reside in the user experience built around that model and in deep vertical integration. The ‘craft’ side of a product will no longer be what the model is, but how frictionless it solves a specific problem.”

I want to shout the result I have gathered: the future will be much more frictionless. User experience designers, do not be afraid! AI will not kill us; instead, it will allow us to deepen our craft.

Agency and Clarity

The long stretch between Rome and Barcelona was more than just a physical distance; it was a space where I silenced the noise to find the “signal” Karpathy talks about. The clarity I reached at the end of those 1000 kilometers is this:

The future will be frictionless, that much is certain. However, in order not to drift away in that smoothness, one needs to fix certain points in their life, anchoring their own agency. This is precisely why I strive to fix certain things to ensure the things that need to move, actually move.

In a frictionless world, accelerating is easy; the hard part is having the internal strength to decide which direction to take.

Thanks Andrej!

We will inevitably see artificial intelligence in every field and every corner of our lives. Concepts like “AI-Supported” or “AI-First” are becoming standard characteristics of every new or existing digital and physical product; this is quite normal.

However, one of the most debated topics is the attempt to squeeze AI into a “human-like or, on the flip side, animal-like” structure by trying to dress it in an “emotional shirt.” Sure, AI can be used for everything to make our lives easier. If we are moving 10 steps a day, AI can carry us x10 steps forward for humanity with its processing power and RAG (knowledge retrieval) capabilities. But what about the other side of the circle: treating AI as an emotional being?

This is something we are still heavily debating. Is it just an innocent curiosity, or is it trying to shift humanity’s perspective on living beings, specifically those with actual emotions and characters, in a different direction? In the end, will it transform our social relationships, whether with animals or humans, into a completely different dimension?

To be honest, we don’t know yet. But it’s clear that serious investments are being made. As always, these experiments happen with animals before humans, similar to how the first living being in space was a dog. While there is significant work on humanoid robots and brain chips, the most intense efforts to “embed emotion into a living form via AI” are happening with animals. I am sure we will see the definitive results within the next five years.

Casio: From My Grandfather’s Watch to an AI Companion

I must say, seeing an “AI Pet” (or as they call it, a “companion”) move from a watch brand really surprised me. In our minds, Casio is the reliable Japanese watchmaker we’ve known for decades. For me, it’s “my grandfather’s wristwatch” brand; it doesn’t exactly scream “innovative.” But Casio is a giant that also produces musical instruments and medical equipment, aiming to be not just on people’s wrists, but in their homes and lives, constantly trying to reinvent itself.

Casio is so determined to enter this space that they want to be part of our lives with a new “Powered by AI” perspective. I see this both in my research and in the way they’ve positioned their new AI companion, Moflin, right in the center of their global website’s main menu. The name Moflin is derived from “Moffu-Moffu” (a Japanese onomatopoeia for the feeling of touching something soft and fluffy) and “Life.”

Moflin is actually a very simple idea, almost a prototype for a giant brand like Casio to release. It has a basic hardware structure: gyroscopes, sound detection, touch sensors, and a simple 1200 mAh battery. Wrapped in a small plush fur, it looks like those old cat figurines. But here is where it gets interesting:

Moflin doesn’t use any popular language models like OpenAI or Google, and it isn’t even connected directly to the Cloud.

Moflin uses an embedded structure Casio calls “Adaptive Emotional AI” (developed in partnership with Vanguard Industries). This is about much more than Moflin being a “cute” housemate; in my opinion, Moflin is an experiment by the brand on producing emotion-based AI products.

• 2D Emotion Map: Moflin’s inner world has a map that shifts across a wide spectrum from happiness to sadness, excitement to calmness.

• Learning Process: It processes data from its sensors on this map. For example, if you treat it gently, it moves to the “happy and secure” zone; if neglected, it may feel “anxious or lonely.”

And the most fascinating part: Casio claims that every Moflin is different. Just like humans or animals... Moflin’s AI learns the owner’s behavior over time to create a unique personality, with over 4 million possible combinations. Instead of massive LLM datasets, it uses a machine learning structure that focuses solely on its interaction with you.

Backing Up a Robot’s Soul

While Moflin isn’t directly connected to the cloud, it connects via its mobile app, MoLife. If you allow it, you can track interaction logs (how much it was loved, when it was left alone) and emotional development data through the app.

The most striking point: if your robot breaks, you can buy a new Moflin and transfer the “soul” (the character developed so far) of your backed-up Moflin directly into the new one.

Vanguard, the company behind Moflin, introduced the concept of “Digital Biology” while designing it. They base this on two pillars:

1. Constant Evolution: Moflin doesn’t just react to fixed commands; it learns every day and develops a “biological rhythm.”

2. Crossing the Uncanny Valley: Moflin’s design intentionally doesn’t mimic a specific animal perfectly. This prevents our brains from asking, “Why isn’t this acting like a real rabbit?” and avoids that creepy feeling.

In Japan, they’ve even launched a subscription service called “Moflin Club,” offering everything from fur replacement to “check-ups.”

Happy New Year to everyone! 🎄 As we leave an intensive year behind, we are starting the first days of the new year with full focus on building and refining.

While we are sharing these updates with you today, please note that these changes are not live yet. We are planning to launch the new versions by the second week of January, and we are continuing to ship new improvements until the very last moment.

Here’s what happened behind the scenes this week:

+ studio updates

Keiki Studio: Clarity and Micro-Improvements

Following last week’s v2.0 upgrade where we pivoted towards an “agentic” logic and a fresh design, this week was all about polishing the user experience.

• New Hero Section: To ensure our value proposition is crystal clear, we added a dedicated section right under the hero area that explains exactly what Keiki Studio does.

• Micro UI Refinements: We implemented several design tweaks to ensure visual harmony across the entire site.

• Roadmap to Launch: We are currently in the final testing phase. While the “Agentic” v2.0 is being shared in these weekly updates, the public launch is scheduled for the second week of January.

+ venture lab {product} updates

Fluxar: Design System Overhaul & Sense Phase Activation

This week was a major technical sprint for Fluxar, focusing on visual consistency and the activation of Fluxar Orbit’s analytical core.

• Design System Overhaul: We have completely redesigned the landing page and user panels. By transitioning to a brand-new design system, we’ve achieved a much more modern and fluid interface.

• Fluxar Orbit™ Sense Phase: Our analytics connection infrastructure is now fully operational. The “Sense” phase has been activated and passed all internal tests. We plan to open this phase to public use very soon.

• Gemini 2.5 Pro Integration: Google Gemini 2.5 Pro is now officially integrated into the Fluxar AI LLM stack. This has resulted in significantly higher reasoning capabilities and more consistent analysis.

• One-Shot Analysis Improvements: We optimized the quick-analysis workflow to provide faster and more reliable results.

• Apiflash Integration: To ensure visual data inputs are as sharp as possible, we implemented high-quality website screenshot capture.

• Demo Mode: We added a dedicated “Demo Mode” that allows our team to showcase all 6 stages of the Fluxar Orbit workflow during presentations and customer meetings.

+ general & investment

On the investment front, there are no new developments at the moment; we remain in a waiting period. We are using this quiet time as an advantage to stay in “deep building” mode and focus entirely on making the product the best it can be.

Happy New Year again, and see you next week!

Thanks for reading keiki weekly! Subscribe for free to receive new posts and support my work.

We spent 2024 in a state of boundless optimism intoxicated by Large Language Models. It was a year where every startup appended .ai to its name, and every problem was presumed solvable by a chatbot. However, 2025 is going down in tech history as the growing pains of leaving adolescence behind. This year, we witnessed software stepping out of its flexible, forgiving nature and crashing into the hard walls of the physical world, while simultaneously unlocking new modes of production we previously couldn’t even imagine.

As product managers, we have claimed for years to be designing frictionless experiences. Yet, 2025 taught us that the real friction lies not in the interface, but in physical reality itself. On one hand, an era of immense ease where coding has transformed into a matter of vibe; on the other, hardware dreams hitting the limits of metal, battery, and energy.

The balance sheet of this year is not just about what didn’t work, but about what has changed irreversibly.

Vibe Coding and the Democratization of Creativity

Before we put on our critical glasses, we must acknowledge the most fascinating revolution of this year: Vibe Coding.

Coined by Andrej Karpathy and sitting at the center of our lives in 2025 with tools like Replit Agent, Cursor, and GitHub Copilot Workspace, this concept has transformed the act of coding from a syntax problem into purely an intent management problem. You no longer need to know Python; you simply need to describe what you want, how it should feel, and what purpose it serves.

This was a historic breaking point for Product Managers. Formerly, to bring an idea to life, we had to convince engineering effort, conduct sprint planning, and wait weeks for an MVP. In 2025, a PM can turn a feature idea conceived over morning coffee into a working prototype via Vibe Coding tools by lunch.

For the first time in tech history, the distance between idea and product has reduced to zero. This is not just a matter of speed; it means the removal of technical barriers standing in the way of creativity. Now, everyone is the architect of their own imagined software.

The Rise of Agents and the Silent Death of SaaS

While the production of software became easier, its consumption changed radically. 2025 became the year AI evolved from a passive assistant into an active Agent.

When OpenAI’s Operator or Microsoft 365’s Agent Mode came online, the equation shifted. Previously, software was designed for humans to use. In 2025, software was designed to do work. Booking a flight, conducting market research, or managing a complex return process... These are no longer tasks performed by a human clicking away at a screen, but tasks delegated to an agent.

This situation plunged the Per Seat pricing model, which has dominated the B2B world for twenty years, into an existential crisis. If one agent can do the work of a 10-person team in 10 minutes, do you sell the company 10 licenses, or 1?

The upward-trending SaaS revenue curves we saw in investor presentations were broken by this productivity paradox. We are no longer selling software; we are selling labor. This means pricing shifts from access to outcome. This is a transformation as frightening as it is exciting; because now we can measure the value of the product we sell not by time saved, but directly by value created.

The Hardware Graveyard and Hard Realities

While such massive leaps occurred in the software world, the physical world told us, not so fast. 2025 became the year Silicon Valley’s hubris of building hardware with a software mindset was most severely punished.

Devices like the Humane AI Pin and Rabbit R1 were marketed as revolutions that would replace the smartphone. The result, however, was devices turning into bricks and disappointment. The lesson here was clear: User behavior changes slower than hardware. When you take screens away from people and give them a heating laser projector strapped to their chest instead, you aren’t solving a problem; you are creating new ones like privacy and ergonomics.

In software, the MVP culture saves lives; you can fix buggy code with a midnight patch. But there is no MVP in hardware. You cannot release a physical product in beta. 2025 reminded us that milliseconds in digital mean damage in the physical world, that battery life is a UX problem, and that the laws of physics do not receive updates.

The Theater of Autonomy and Teleoperation

The same hard reality confronted us in the world of robotics. It turned out that behind those smooth movements we saw in Tesla Optimus and 1X Neo videos, there were often human operators wearing VR headsets.

Is this a failure? I think not. This is proof of how difficult the process of technological embodiment is. AI hasn’t fully embodied yet, but a new symbiotic relationship has been born between human and machine. We provide the mind; they provide the muscle power. However, a major risk arose for product management: Latency. When a chatbot answers late, you get annoyed; but when a remotely operated robot reacts late, it breaks the glass in its hand.

The Bill for Intelligence: Watt, Water, and Dollar

Right behind the hardware crises, another physical cost of the digital world emerged: Energy.

We were accustomed to thinking of software as infinite and free. But 2025 marked the end of this illusion. As AI models got smarter, their hunger for energy increased geometrically. It was no coincidence that Google and Microsoft accelerated their nuclear energy investments this year.

For a product manager, this means adding a new metric to the dashboard: Inference Cost. Previously, server costs were a negligible detail. Now, every single reasoning step of an autonomous agent translates into a real electricity bill and water consumption.

In 2026, the difference between a premium product and a standard product will be determined not by features, but by the intelligence level of the model used and the energy it consumes. For the wealthy, models that think deep and consume high energy; for the masses, models that are fast, shallow, and cheap. The digital class divide is now drawn not by access to data, but by access to energy.

PM 2.0: The End of Backlog Management

So, what happens to the Product Manager in this world of agents, autonomous coding, vibe coding, and energy constraints?

Let’s be honest: If your job is moving tickets in Jira, taking meeting notes, and grooming the backlog, 2025 was tough for you. Because AI agents now do these tasks better, faster, and 24/7. However, this is not the end of the PM; it is their renaissance.

Freed from the burden of being an executioner, the PM is returning to their true role: the System Architect. The successful product person of 2026 is the one concerned not with how it will be done, but with why and what will be done. As writing code becomes cheaper, choosing what to build becomes more expensive.

Human intuition, strategic vision, empathy, and ethical decision-making... These are competencies not found in an LLM’s training set, unable to be converted into tokens. The PM is no longer a project manager, but an orchestra conductor managing the digital workforce at their disposal.

Techno-Geopolitics and Digital Minimalism

When we look at the whole picture, we see that technology is no longer a sterile laboratory product. The destructive price war launched by China’s DeepSeek against OpenAI and Europe’s defense of Sovereign AI with Mistral proved that code now has a nationality.

And amidst all this noise, human nature developed its own defense mechanism: Digital Minimalism. It is no coincidence that devices which disconnect, like the Light Phone III and reMarkable Paper Pro, broke sales records in a world where everyone is trying to sell more connectivity. People no longer want supercomputers; they want Calm Tech. Those rare moments where notifications are silenced and attention is undistracted became the most luxury consumer product of 2025.

Conclusion: The Easy Days Are Over, The Meaningful Days Have Begun

Closing 2025, we are far from the blind optimism of 2024, but we stand on much more solid ground.

We now know that AI is not a magic wand; sometimes it’s an intern writing great code, sometimes an assistant handling your tasks, and sometimes a child stumbling in the physical world.

The winning products of next year won’t just be those training smarter models. The winners will be those who integrate that intelligence into human life with the least friction, use energy efficiently, and most importantly; respect human attention and intent.

My 2026 Prediction: The Compression of Time

As I finish this piece, I want you to set aside product strategies and technical details for a moment and look at the calendar. Because that dizzying feeling we have is not psychological; it is entirely mathematical.

Let’s freeze the historical perspective for a moment: It took humanity approximately 1900 years to get from agriculture to the Industrial Revolution. From there, jumping to electricity and the internet took us only 60 years. The transition from Web 1.0 to Web 3.0 fit into 20 years. And moving from Generative AI to the era of Agents and Robots we discuss today? Only 2 years.

The fabric of time is compressing. Innovation cycles are no longer measured in generations or decades, but in months.

Therefore, my prediction for 2026 is clear: The coming year will be a faster year than ever experienced in the history of technology. We will witness the time it takes for the concept called future to turn into now approaching zero. Fasten your seatbelts; because the cracking sounds we heard in 2025 were not of destruction, but the sound of a rocket taking off.

Welcome to the fourth edition of keiki weekly. This week marks a significant milestone in our digital presence. We have transitioned from a static showcase to a dynamic, agentic ecosystem while simultaneously pushing the boundaries of Fluxar’s core intelligence.

+ studio updates

1) keiki Website v2.0 & Agentic keiki

We have completely overhauled the keiki web experience to reflect our evolving identity. This update moves beyond simple aesthetics into functional intelligence.

• Website v2.0 Reveal: A brand new interface and refined user experience that aligns with our core design philosophy.

• Agentic keiki Integration: We have integrated the keiki Agent directly into the site. You can now interact with the system to learn everything about the keiki Venture Lab and our AI Solutions Zone.

• Refined Feel: The entire site atmosphere has been updated to provide a more immersive and technically fluid experience.

  visit keiki web

2) Investment & Partnership Developments

The interest in Fluxar continues to expand within the global venture capital landscape.

• Global Investor Interest: Another international investor has reached out to the studio regarding Fluxar. We are currently moving forward with active discussions to explore strategic alignment and future growth.

+ venture lab {product} updates

Fluxar: Strengthening the Core and Sensing the Data

This week’s technical sprint for Fluxar focused on model reliability, visual precision, and the first steps toward autonomous data sensing.

• One-Shot Analysis Improvements: We have enhanced the overall analysis experience, providing more accurate and actionable results for every query.

• Gemini Model Integration: Google Gemini 3.5 Pro is now officially connected to the Fluxar AI LLM stack, resulting in more consistent and high-reasoning responses.

• Apiflash Integration: Implemented high-quality website screenshot capture to ensure visual data inputs are as clear as possible.

• End User Panels: Refined the user dashboard and panels to improve navigation and ease of use for our early adopters.

• Fluxar Orbit™ Sense Phase: We have successfully built the analytics connection infrastructure and initiated the first phase of testing.

Thanks for reading keiki weekly! Subscribe for free to receive new posts and support my work.

Look back at the last few years. AI has shifted from a niche topic at tech conferences to the constant companion of our daily work. Regardless of our role in the digital product world, whether we are product managers, developers, or designers, it is now impossible to imagine a product vision without AI. When planning a feature, optimizing code, or designing a user experience, we all ask the same question: “How can we make this better with AI?”

AI is no longer a luxury. It is a utility, like electricity for digital products. But as a Builder, I believe we have to question where and how we source this electricity. We usually choose the easiest path. We connect to a massive API, send the data, and get the answer. But what happens behind that magic box?

To ask these questions, I sat down with , the founder of GreenPT.

Blurple Studio, B Corp, and Shared Values

My connection with Robert actually goes back about four months. It was during the exciting period when Blurple Studio, the UX studio I founded, received its B Corp certification. For those who know, B Corp is not just a certificate. It is a stance. It is a commitment to using business as a force for good, not just for profit.

Connecting with Robert was no coincidence because he is trying to do exactly the same thing in the AI world. He is blending technology with ethics and sustainability.

So, What is GreenPT?

Before diving into the interview, let’s set the context. Calling GreenPT “just another AI tool” would be unfair. GreenPT is a privacy-first, sustainable AI platform designed for enterprise, hosted entirely in Europe.

What keeps it on my radar as a developer is that it is not just a wrapper. GreenPT is a full ecosystem that serves open-source models on its own optimized infrastructure while transparently showing where your data goes and how much energy it consumes.

Robert and I did not actually meet in the sunny, green office shown in the image above. I was in Barcelona, and he was in The Netherlands. However, we asked an AI to imagine the transparent environment our conversation deserved. In that virtual setting, we discussed very real problems.

How It Started: “A Practical Frustration”

I started by asking Robert about the initial spark for GreenPT. He did not give me a philosophical speech. He described a practical engineering frustration.

AI adoption was moving fast, but the default setup was broken. It was usually US-hosted, opaque, energy-heavy, and hard to trust.

“I wanted a platform that makes the responsible choice the easy choice,” Robert told me.

His vision was clear. He wanted European hosting, privacy by design, and open technology that prevents vendor lock-in. Most importantly, Robert realized these values are not decorations. “These are not separate features,” he noted. “In real adoption, you cannot bolt them on later without compromises. They must be in the product DNA.”

Sustainability is a Stack, Not a Sticker

I asked Robert how the sustainability aspect of GreenPT works in practice. His answer showed that it is not about buying carbon offsets to clear your conscience. It is a stack of choices.

1. Renewable Energy: Picking infrastructure partners and regions, specifically in Europe, where energy sourcing is demonstrably cleaner. They also keep workloads close to users to reduce network overhead.

2. Efficient Model Selection: Instead of using heavy models for everything, the default is to use smaller, strong open models for everyday tasks.

3. Optimization: They use quantization, caching, and smart context handling. If a shorter context works, the system does not reprocess huge prompts.

4. Operational Guardrails: The platform steers users toward “good enough” and efficient outputs instead of always using maximum compute.

A Privacy-First Architecture

In the B2B world, trust is an architectural requirement, not a marketing slogan. So I asked about their privacy design. Robert’s answer relied on a single principle: Data Minimization.

“Do not collect what you do not need, and do not store what you cannot justify,” he said.

This principle translates into concrete decisions at GreenPT:

• European Hosting: For predictable jurisdiction and GDPR alignment.

• Transparent Data Handling: Clear rules on what is logged and what is retained.

• Zero Tracking: A strict product stance.

Robert emphasized that if you want organizations to use AI in real workflows, the architecture must stand up to security and compliance scrutiny without vague excuses.

Open Source and the 90% Rule

One of Robert’s most striking observations concerned the efficiency of open-source models. In an era where everyone is chasing the largest proprietary model, Robert argues that we are over-computing.

“90% of AI tasks can be handled by open models,” he stated.

He sees this in daily work. Summarizing, rewriting, searching a knowledge base, drafting, or simple classification. For these tasks, open models are often better than the giants. “The advantages are huge,” Robert said. “Lower cost, more control over data, deployment flexibility, and fewer compliance headaches.”

Why a “Full Stack” Platform?

I asked why he describes GreenPT as a full stack platform rather than just a chat interface. He explained the layers involved:

• Chat: For human-friendly interaction.

• API Access: So teams can embed GreenPT into apps and internal tools.

• Model Routing: To ensure the right model is used for the right job.

• Real-time Insights: To measure CO2 and energy usage, making sustainability a feedback loop inside the product.

The Roadmap and Future Goals

What comes next? Robert shared the roadmap in two parts. Product-wise, they are building better routing between open and premium models, stronger governance controls for organizations, and more connectors for internal data retrieval.

Movement-wise, they are pushing for shared standards around measuring AI energy impact. They want to help buyers ask better questions, such as “Where is this hosted?” and “Can I switch providers?”

The Trust Gap and PitchAI

Finally, I asked about the core problem they are determined to solve as AI adoption grows. Robert identified the biggest gap as trust with accountability. Many tools are impressive but unclear on data handling. GreenPT aims to close that gap by making privacy and transparency operational defaults.

We ended our chat by discussing PitchAI, a recent project where GreenPT helped younger participants refine business pitches using AI. It was a perfect proof of concept. It provided a safe environment where young creators could explore AI without tracking or data ambiguity.

As I finished the call and returned to my own development work, the lesson was clear. Efficiency and privacy are not barriers to innovation. In fact, they are the only way to build a better, more reliable digital product.

Welcome to the third issue of keiki weekly.

This week we are focused entirely on product velocity. We are expanding the vision for Fluxar by initiating its true form, “The Orbit,” and we are launching a new standalone product, Houra AI, to streamline our workflows.

Keiki continues to grow with the same principle: AI-native creation, human-led direction.

✦ venture lab (product) updates

1) Fluxar: From One-Shot Analysis to “Fluxar Orbit”

Until now, the public version of Fluxar has served as our showcase. It was the training ground where we built Fluxar’s “AI Brain” capable of analyzing websites and screenshots.

Now, we are starting to build what we originally set out to create: Fluxar Orbit™.

Orbit is our Continuous Improvement System. It transforms Fluxar from a tool that just “spots errors” into a proactive partner that continuously watches and optimizes your product.

Here is the 6-stage cycle that drives this engine:

1. Sense: Collect live analytics data (GA4, Clarity, Hotjar).

2. Diagnose: Run AI analysis validated by expert insights.

3. Orchestrate: Estimate impact and prioritize the roadmap.

4. Prototype: Generate AI solutions (UI copy, visual design, components).

5. Measure: Track performance through A/B testing results.

6. Loop: Feed learnings back into the Experience Memory™ system.

This Week’s Focus: Phase 1 (Sense) & Infrastructure

We are currently building the “Sense” phase to feed this cycle with real data.

• User Panel & Persistence: We are moving beyond guest access. Users can now sign up to save, track, and manage their reports.

• Analytics Integration (Sense): The first leg of Orbit is active. Users can connect their own analytics tools to Fluxar. This allows for data-driven, sustainable analysis outputs based on real user traffic.

• Human-in-the-Loop: We added a critical layer. Users can now request comprehensive audits from UX Experts directly through the Fluxar panel.

• Gemini-Powered Prediction: We integrated a Gemini-based structure into the analysis engine. This provides highly consistent, heatmap-based prediction analyses for websites and screenshots.

• Tech Stack: We are migrating our database infrastructure to Supabase for better scalability.

2) Houra AI: Your New Calendar Assistant

We also shipped a completely new product this week. Houra AI is now live on Slack.

Houra removes the friction from scheduling. By adding the Houra Bot to your Slack workspace, managing your calendar becomes conversational.

• Natural Language: Just mention the bot to take action.

  • Example: “@houra schedule a 30-minute meeting with Deniz this week.”

• Smart Sync: Houra analyzes both parties’ Google Calendars, identifies the best available slot, and sends the invite automatically.

We are building an ecosystem where data, AI, and human expertise work in sync.

See you next week as we continue building the Orbit.

Thanks for reading keiki weekly! Subscribe for free to receive new posts and support my work.

Thanks for reading keiki weekly! Subscribe for free to receive new posts and support my work.

Welcome to the second issue of keiki weekly.
This week, we’re sharing structural improvements across keiki studio and a major technical sprint on the product side.
Keiki continues to grow with the same principle:
AI-native creation, human-led direction.

✦ studio updates

1)keiki Hub Updates

This week, we strengthened both the information architecture and the decision-making layer of Keiki Hub.

• Product roadmap is now vote-enabled: Community input becomes more visible and influential.

• Information architecture refreshed: Products, documentation, and content reorganized into a more intuitive structure.

• Navigation and content flow improved: The branches of Keiki are now easier to follow and understand.

Keiki Hub is becoming the living map of our studio. let’s discover: https://www.keiki.studio/hub

2)Investment & Partnership Developments

An important milestone for Keiki this week:

📍 A Barcelona-based venture capital firm reached out to us, expressing interest specifically in our venture studio model.

The process is currently under NDA and at a very early stage.
We’ll share updates as soon as we can.

A promising signal of Keiki’s growing visibility beyond local boundaries.

3)Publication Layer: New Interview Published

This week, as part of keiki studio’s publication layer, we released a new interview on product.blog:

ADPList – A Conversation with Felix Lee on Mentorship, Community, and Building for Real Impact
🔗 https://product.blog/adplist-a-conversation-with-felix-lee-on-mentorship-community-and-building-for-real-impact-b871c8b78eb3

A thoughtful conversation on mentorship, community, and designing for meaningful impact.

✦ venture lab (product) updates

Fluxar: This Week’s Technical Sprint

Fluxar underwent a major technical upgrade this week.
All updates from the sprint expand Fluxar’s capability set:

• Puppeteer integration: Secure access to masked sites via browserless.io.

• 500 endpoints added to the analysis pipeline: Expanded data streams for richer pattern detection.

• Admin panel refreshed: Faster, cleaner operational visibility.

• Claude added to the analysis pipeline: Fluxar now runs on a dual engine—GPT + Claude.

• Improved GPT search & notes system: More consistent chat results and normalized note structures.

• Product Interaction Room launched: A live web space to observe and understand product behaviors and flows.

Fluxar didn’t just grow this week, it became more intuitive and more insightful.

Let’s try new version of Fluxar ➡️ https://fluxar.ai/

This week strengthened both sides of the keiki ecosystem:
Structure on the studio side, engines on the product side.

The roots are deepening, the branches extending.
More updates from a different part of the ecosystem next week.

Thanks for reading keiki weekly! Subscribe for free to receive new posts and support my work.

In the early months of COVID, that luck disappeared. The hallways emptied. The conversations stopped. People were losing their jobs while the world held its breath, and the one thing many needed most guidance suddenly slipped out of reach.

I reached out to Felix Lee to understand how ADPList emerged from that moment of uncertainty. This piece is shaped around the written responses he shared with me, and the story behind them reveals a product built not from opportunity but from necessity. ADPList is a platform that connects people with mentors across design, product, and tech, making high-quality guidance accessible to anyone around the world.

How It Started

When I asked him what originally inspired ADPList, he answered with a clarity that usually signals a lived problem rather than an abstract idea. His response cut straight to the heart of it.

“People were laid off during COVID and couldn’t find mentorship,” he said.

There was no dramatic buildup, no polished founder narrative. Just a simple truth about people in need of guidance, unsure where to turn. For him, this wasn’t a conceptual or strategic challenge. It felt immediate and human. He wanted mentorship to be reachable for everyone, everywhere. That intention became the foundation for everything that followed.

A short time later, he made the decision that would define the platform.

“Making mentorship completely free,” he told me.

It was a simple call with unexpectedly large consequences. Mentors joined because they genuinely wanted to contribute. Mentees arrived without hesitation. Removing the price tag created trust faster than any growth tactic could. That single decision set the cultural tone and shaped ADPList’s trajectory.

Building a Community by Letting It Breathe

When I asked how ADPList grew into such a broad and surprisingly positive community, his response was almost understated in its simplicity.

“Authenticity and giving first,” he explained.

There was no engineered mechanism behind the growth. No aggressive playbooks. People had meaningful conversations and shared them with others. Companies discovered ADPList in exactly the same way. The product moved through individuals, not campaigns.

Curious about how they maintain this experience at scale, I asked which signals the team pays most attention to. He continued with the same thread.

“Quality conversations. Repeat sessions. Match quality. Feedback. That’s what we focus on,” he said.

ADPList was never intended to be a massive directory. It was built to make the right conversations happen. The metrics they track reflect that intention: depth over volume, impact over activity.

Where AI Fits Into Mentorship

In 2025, any product conversation eventually finds its way to AI. I asked how he sees AI shaping the future of mentorship and how it already fits into ADPList today. His answer was thoughtful and measured.

“AI helps with matching and discovery. We’re exploring how AI can enhance, not replace, human mentorship,” he said.

AI, for him, is a supporting layer rather than the centerpiece. This approach aligns with the thinking in his “Design by AI” ebook, which frames AI as a collaborator that expands creative possibility instead of flattening it.

Across all his answers, one theme returns consistently: technology should lift the human connection, not overshadow it.

When Companies Join Through Their People

ADPList’s homepage features companies like Booking, Notion, Spotify and Figma. I asked how these relationships began, and his explanation reflected a pattern already visible in the platform’s growth.

“They came organically,” he said. “Employees used ADPList, loved it, and brought it into their companies for internal mentorship programs.”

No partnership pipeline. No outbound effort. Just individuals who found value and carried it into their teams. In many ways, this mirrors how ADPList spread publicly. Trust moved quietly at first, then consistently.

Before we wrapped up, I asked him if he had any final thoughts for readers of product.blog. He paused, then offered a single line that captured everything he had expressed throughout our conversation.

“Build for impact, not just growth. When you solve real problems for real people, everything else follows,” he said.

ADPList didn’t begin with a plan for scale. It began with a need and a decision to meet that need. In a world crowded with products competing for attention, his story is a reminder that meaningful things often begin quietly, with a simple intention to help.

Check out ADPList: https://adplist.org/

ADPList: A Conversation With Felix Lee on Mentorship, Community and Building for Real Impact was originally published in product.blog on Medium, where people are continuing the conversation by highlighting and responding to this story.

In botany, a new life that grows from the stem of an orchid, carries the same genetic code, yet anchors itself by growing its own roots, is called a “keiki.” This word goes beyond botany. In Hawaiian, it means “child.” In Japanese, it represents luck and opportunity. Wherever you go in the world, “keiki” stands for renewal and sustainable beginnings.

Thanks for reading keiki weekly! Subscribe for free to receive new posts and support my work.

Our story began inside the blurple ecosystem through the same kind of natural evolution. Over the last two years, our relationship with AI has transformed not only the way we work but also the way we think. What grew inside us eventually became too large to stay contained and turned into something independent.

Today, that shoot is being planted into its own soil. Its name is keiki studio.

We are a next-generation structure born with AI, centered on technology, and guided by people.
We believe production today requires a hybrid mindset.

Hybrid production discipline

We are at a breaking point where technology is becoming increasingly complex, while answers are becoming more accessible. But we believe the future lives not only in lines of code, but in the human intention behind them.

For us, AI is not a replacement. It is a lever. Being AI-native does not mean handing over thinking to machines. It means extending human intelligence with machine speed and building a hybrid consciousness through the right interaction between humans and AI.

We bring this philosophy to life through two core structures that feed into each other:

1.Our kitchen: AI Venture Lab

This is our production base as a venture studio, where we turn our own ideas into real products. We do not focus on a single layer of technology.

End-to-end products: Platforms designed from the ground up, using AI as the core engine to solve complex problems at scale.
Micro software solutions: Agile tools that focus on specific problems and deliver fast, smart results through built-in AI capabilities.
Experimental projects: Prototypes that explore new use cases and test the boundaries of what AI can do.

2.Our collaboration space: Dev & Research Zone

Keiki is also a door of opportunity, just as its meaning in Japanese suggests. As we grow our own ecosystem, we also support others in creating their own keikis.

Who is this for?
Founders building their own products, companies improving workflows with AI, and teams looking for technical guidance and experience.

The blurple difference:
We do not only write code. We combine our engineering strength with the deep experience design capabilities of our strategic partner, blurple studio. This allows us to go far beyond what is usually expected.

When knowledge is shared, it takes root: our culture of open learning

We are not building behind closed doors. We move forward in the open. With an “open development” mindset, we created two publishing spaces to document and share our journey:

product.blog: Where we tell the full stories behind our digital products, whether AI-related or not.
keiki weekly: A weekly newsletter where we share what we learned, what we built, what challenged us, and how we moved forward, without filters.

If you’d like to walk with us, exchange ideas, or create your own keiki, our door is always open.

We were born with AI. We grow with people. Here’s to many keikis ahead.

Thanks for reading keiki weekly! Subscribe for free to receive new posts and support my work.