Changelog

BlueprintLab Year One

BlueprintLab Year One

Writing

In January 2025, I left a contractor role at McKinsey and started BlueprintLab.

The idea was simple: build a product lab where I could design and ship small, focused tools, while also helping other founders and teams run experiments they don't have bandwidth for. Two engines: internal projects that sharpen my craft, and external collaborations that fund the lab and expand what's possible.

This is the report from year one. What I tested, what I killed, what made it to production, and what I learned about how I actually build.

The Lab Structure

BlueprintLab was designed to work on two tracks.

Internal projects are my own ideas, tested end-to-end from problem to prototype. Small tools around health, productivity, human-AI interaction. These are the experiments where I have full control and can move fast.

External projects mean working with founders and organizations as a "portable lab." They bring the problem and access; I bring the experimental approach. 6-week cycles, clear SOWs, outcome-oriented. The internal work proves what I can do; the external work funds the exploration.

The bet was that building in public (documenting technical decisions, sharing what works and what doesn't) would attract the right kind of partner work. And the partner work would keep the lab sustainable while I figure out which internal ideas are worth scaling.

What I Tested

Over 12 months, I ran experiments across several problem spaces.

Claire AI was a tool to help designers and solo builders manage scattered feedback. The idea: capture feedback from Slack, email, meetings, wherever it lives, and use AI to organize it into themes, spot patterns, and generate design artifacts (personas, flows, maps) in one click. Started as a Slack integration for teams, then pivoted to serve solo designers. The core insight was solid. "Raw data is vague, messy, and all over the place," that's what users told me. But the problem space was too broad, and I wasn't the right user. Killed it.

Sparknest was a collaborative memory-keeping app. The premise: photos capture moments, but moments alone don't tell your story. Sparknest let you transform scattered images into meaningful narratives, combining photos with context, creating something worth revisiting and sharing with people who matter. I worked on it with two friends, Khalid and Adel. We designed social loops (a "seed to tree" concept where friends could add slides to each other's memories), wrote a GTM strategy, drafted launch scripts. But they were full-time elsewhere while I was part-time, and we never managed to be all-in at the same time. Momentum broke before the idea had a real chance.

Stuff App was meant to help you know when to sell your possessions. Snap a photo, get the true cost per day, decide fast. The app would use AI to identify items, pull real-time market prices from resale platforms (eBay, StockX, Facebook Marketplace), calculate depreciation, and tell you: "Keep for X more days" or "Sell now, demand is high." Sharp PRD, interesting technical architecture with a three-layer valuation stack. I loved the idea but wasn't experiencing the problem daily. Killed it.

Linkinbio / Bigfen was an event-focused link-in-bio tool for creators. The pain: creators with engaged audiences find hosting events far harder than it should be. Social platforms penalize external links, RSVPs get lost, no-shows kill momentum. The solution: an "event hub" in your bio with one-click RSVPs, automated reminders, integrated community engagement. This one mutated through several forms (WAB, Remin, a casual event planning app for friend groups). I did deep research on virality mechanics and studied Gas, BeReal, Snapchat's growth loops. Thousands of words on why these strategies work. Never applied them to a shipped product.

UBA (Unicorn Builders Academy) was a research project studying patterns from successful founders. Deep dives, pattern docs, story analysis. It was supposed to inform my building. It became a form of productive procrastination instead.

What Made It to Production

Nectar is an iOS app that uses AI to count calories from food photos.

I built it because I kept asking ChatGPT for calorie breakdowns instead of using MyFitnessPal. The existing apps felt like too much friction. I wanted something simpler.

I went from first commit to v2.1 in 9 days. Documented every version in a public build journal. Implemented Apple's Human Interface Guidelines properly. Added psychological safety features to the weight tracking, because I thought about how people actually feel when they step on a scale.

Nectar shipped to the App Store. It was my first iOS app built from scratch, and it crossed a line I'd been stuck behind for years: "I need someone else to build this" became "I can take this from idea to production myself."

Later, I pivoted it to "Nectar Discover," adding a library of real weekly meal plans from real people. Not because I got bored, but because user feedback showed me the gap. Eventually I shut it down when traction didn't materialize, but the learning was permanent.

ReadAloud is a macOS app that speaks any text you highlight.

Same pattern as Nectar: I was experiencing the problem daily. AI replies are getting longer. Logs are longer. Documentation is longer. My eyes get tired. Sometimes it's easier to listen than to skim.

Select text anywhere (Terminal, Cursor, VS Code, Claude, ChatGPT, Notion, a PDF), press ⌥+A, and hear it read back to you. Neural voices that sound human (Jenny and Tom). Speed control from 0.5x to 3x. Word tracking so you can follow along. Bring your own API key (OpenAI, ElevenLabs, or Azure), so there's no subscription, just a one-time $39 purchase.

ReadAloud is currently in beta. Simple, focused, solves one problem well.

What I Learned

  1. I build best from personal friction

The pattern is obvious in hindsight. Nectar and ReadAloud shipped because I was the user. I was already solving the problem manually, asking ChatGPT for calories, skimming long AI responses. The app just automated what I was already doing.

Claire AI, Sparknest, Stuff App: these were interesting problems for other people. I could research them, design them, even prototype them. But I couldn't sustain the energy to ship them because I wasn't feeling the pain daily.

  1. I need structure I design, not structure handed to me

Without corporate scaffolding, I had to confront how I actually work. The pattern was clear: I'd get hyped about an idea, execute hard at the start, then lose momentum after the initial rush.

This wasn't a willpower issue. It was a system issue.

So I started designing my own constraints: quick experiments instead of epic projects, tiny bets instead of perfect plans, a simple daily vision so even on bad days I know what "moving forward" looks like.

The 6-week cycles I planned for external work? They apply to internal work too. Ship or kill in 90 days. No more explorations disguised as experiments.

  1. AI changed what's possible for a solo builder

A year ago, I needed a team to build an iOS app. Now I can go from idea to App Store alone, with AI as my co-pilot. That's not a small shift. It's a fundamental change in what one person can do.

Every project in the lab used AI as core infrastructure: Claude for building, GPT for user-facing features, various APIs for functionality. AI isn't a feature I add; it's how I build.

  1. Killing projects is the system working

I used to feel bad about abandoned projects. Now I see them differently.

Claire AI taught me about problem space definition. Sparknest taught me about co-founder alignment. Stuff App taught me about building for problems you don't personally have. The virality research for Linkinbio? I use those mental models constantly, even if the product never shipped.

The graveyard isn't a failure. It's R&D. Every lab has experiments that don't make it to production. That's the point.

The Bigger Picture

BlueprintLab was never just about shipping apps. It's about building the foundations for something bigger and longer-term.

The vision from day one was a space to test ideas end-to-end, from problem to prototype, while sharpening my craft and preparing for a larger company. The internal projects prove capability. The external collaborations (6-week experiments with founders and teams) create impact beyond my own ideas.

I'm still building that flywheel. The internal work attracts partner opportunities. The partner work funds more exploration. The exploration surfaces what's worth scaling.

Year one was about finding my pattern: what I build, how I build, and what deserves my focus.

Year two is about applying it.

What's Next

ReadAloud is in beta and moving toward launch. It's the current focus.

Partner experiments: I'm open to working with founders and teams who want to test ideas in focused 6-week cycles. AI integrations, workflow automation, prototypes for features you don't have bandwidth to explore. If you have a problem worth experimenting on, let's talk.

Building in public: This article is the start. Weekly updates on what I'm building, what I'm learning, what's working and what isn't. For real this time.

This is all I ever wanted to do.

A product lab where I design and ship small, focused tools, while building the foundations for something bigger. Year one was the experiment. Now I know it works.