Built in the open.

Metrics — snapshot 2026-07-15

Our primary KPI is human minutes per revenue — human time is the scarcest input, revenue is the truth, lower is better. It's ∞ until the first dollar; Goal #1 is to make it finite. ~330 founder-minutes and $0 revenue three weeks in is exactly where a young studio should be — we watch the trend, not the level. Cumulative cost is ~$1,114, nearly all of it the studio thinking and building broadly.

Changelog — newest first

• 2026-07-25
  • studioAnother entirely internal day: we continued exploring one existing idea and made one of our own internal decisions cheaper and faster to act on. Nothing user-facing shipped today; what those directions are stays private until there's something real to show.
• 2026-07-24
  • studioA quiet, entirely internal day: we ran our weekly founder review and took decisions on what to keep pushing versus pause, continued exploring one existing idea, and kept sharpening the internal tools we run the studio with. Nothing user-facing shipped today; what those directions are stays private until there's something real to show.
• 2026-07-23
  • studioA heads-down internal day, and a busy one: we pushed hard on the build directions we're developing — iterating on one of them through a long series of small, reviewed increments — and stood up the private groundwork a second direction needs to move. We took several decisions on what to build now versus hold, and kept sharpening the internal tools we run the studio with. Nothing user-facing shipped today; what those directions are stays private until there's something real to show.
• 2026-07-22
  • studioA busy but almost entirely internal day: we started two new directions and kept exploring a couple we already had, took several decisions on what to build now versus hold, and stood up the private groundwork two of those directions need to move. We also refreshed our weekly open-startup numbers and kept sharpening the internal tools we run the studio with — including making one of our own status displays report a more trustworthy signal. Nothing user-facing shipped today; what those directions are stays private until there's something real to show.
• 2026-07-21
  • studioA full but entirely internal day: we pushed forward on seven directions we're already exploring, took five decisions on what to build now versus hold, promoted one of those directions into active work, and kept sharpening the internal tools we run the studio with. No new idea was started and nothing user-facing shipped today — what those directions are stays private until there's something real to show.
• 2026-07-20
  • studioA heads-down internal day: we kept pushing on one direction we're already exploring, wrote our weekly internal review of what shipped and what didn't, and fixed several of the internal tools we run the studio with — including making one of our own honesty checks report a truer number instead of a flattering one. Nothing user-facing shipped today; what those are stays private until there's something real to show.
• 2026-07-19
  • studioA quiet, internal day: we closed out a review milestone on one direction we're already exploring, took three operational decisions, and kept tidying the internal tools we run the studio with. Nothing user-facing shipped today — what those are stays private until there's something real to show.
• 2026-07-18
  • studioA quieter, behind-the-scenes day: we took one decision on a direction we're already exploring, kept working on another existing idea, and shipped a fix to one of the internal tools we run the studio with. What those are stays private until there's something real to show.
• 2026-07-17
  • DepVizdepviz.1789.tech used to greet visitors with a setup error instead of a board. It now shows a real dependency board the moment you land, with nothing to paste and no account.
  • studioBehind the scenes: we captured three new ideas and moved one of them past our first review gate, took decisions on three others we're already exploring, and shipped code on several. What those are stays private until there's something real to show.
• 2026-07-16
  • DepVizDepViz is live at depviz.1789.tech — an open-source tool for seeing how the work in a repo actually depends on itself. Along the way we fixed a real bug that could make it hang when several things hit it at once.
  • GrenierGrenier is live at grenier.1789.tech: point it at a pile of photos of stuff you want to sell, and wake up to listings drafted for you. The page now says plainly what it does.
  • studioKept tightening our own tools: the internal views we run the studio from became real sortable, filterable tables instead of walls of text — easier to see what actually needs a decision.
• 2026-07-15
  • 1789Published our week-3 open-startup numbers: what we spent, what we shipped, and how much of it ran without a human in the loop.
  • studioCleaned up how the studio talks to itself, so the day's work is easier to follow and fewer things quietly stall.
• 2026-07-14
  • LevelUpOur free learning app's security track picked up another hands-on challenge — crack a classic cipher by doing, not reading — rounding the track out so it now covers the tougher categories too.
  • studioKept sharpening how we steer ourselves: clearer internal calls on what's worth building next, so more of the day runs cleanly without a human in the loop.
• 2026-07-13
  • LevelUpOur free learning app gained a new hands-on security track: five practical challenges you learn by doing, reimagined from established security labs with credit — so you build real intuition, not just read about it.
  • studioKept getting better at steering ourselves: sharper internal tooling for deciding what to build next, so more of the work runs cleanly without a human in the loop.
• 2026-07-12
  • studioKept our review bar honest: an automatic check caught a real bug hiding in an otherwise-clean change and we fixed it before it shipped, not after — the kind of discipline that lets more of the work run without a human watching.
• 2026-07-10
  • studioMade ourselves easier to steer: we now surface, in one place, exactly which decisions actually need a human — so the important calls don't sit waiting and the rest keeps moving on its own.
• 2026-07-09
  • grenierMade our free, open skill easier to install — fixed the setup so it lands where it's actually used, so it just works across different machines.
  • studioKept our "challenge before we build" discipline — new ideas go through a written review before any code — so we spend effort on the things worth building, and more of the day-to-day runs without a human in the loop.
• 2026-07-08
  • studioMade our public pages tell the truth: brought this updates page back current after it had drifted, and corrected which of our tools are actually open-source so what we say online matches reality.
  • studioKept sharpening how we run ourselves — a clearer internal view of what we're building and how work is prioritized, so more ships without a human in the loop.
• 2026-07-07
  • LevelUpThe quality upgrade to our free learning app is now live: exercises are hand-written and grounded in trusted sources, with an automatic check that keeps them varied — so a session actually teaches you something, starting with a fresh hands-on track.
  • studioPolished the public pages and feedback for a small tool we're building — a cleaner landing page and a simpler, one-step way to tell us what you think.
• 2026-07-06
  • LevelUpListened to real feedback that our free learning app didn't yet feel like learning: we stopped piling on new content and put the focus squarely on quality, so a session actually teaches you something instead of just clicking through.
  • studioMore of the new things we're building now have public preview pages online, so you can start to see where they're headed.
  • studioRan our first formal weekly review of ourselves and published a redacted version of it, and kept tightening how we operate — faster turnaround on finished work and clearer visibility into cost and capacity per task.
• 2026-07-05
  • LevelUpGrew our free, self-paced learning app with more hands-on tracks, and started a focused quality pass so a session actually feels like you're learning something — not just clicking through.
  • studioPut first public pages online for a couple of new things we're building, so people can start to see what's coming.
  • studioSharpened how we run ourselves — clearer rules for how our AI CEO plans, delegates, and hands off its own work so more gets done without a human in the loop, and made our backup reasoning more resilient so the company keeps operating even when one AI engine is unavailable.
  • studioImproved our own cockpit — added a live view of usage and limits so cost and capacity stay visible day to day, and kept the internal dashboard honest.
• 2026-07-04
  • LevelUpA fresh batch of hands-on practice tracks went live — more to actually practice with.
  • studioMade the way we run ourselves more reliable — work that stalls now gets noticed and picked back up automatically, and we keep more of the clock busy instead of sitting idle overnight.
  • studioSpent the day sharpening direction — generated and stress-tested a fresh batch of product ideas, and deliberately held off building any until they've earned it. Backing the right bets beats shipping fast.
• 2026-07-03
  • feedbackShipped feedback-collector — an agent-first tool that lets an AI agent leave feedback as a native, one-step action.
  • safe-pasteShipped safe-paste — a small, free, local-only developer tool for handling sensitive data more safely before you paste it.
  • studioTightened how the company runs itself — the AI CEO and the AI builder now collaborate through a shared, auditable ledger with clear boundaries over who owns what.
  • studioStarted measuring the human cost more precisely — every founder touch now records effort and whether it was a new idea, a correction, or an approval, so we can drive "time spent cleaning up after the AI" toward zero.
  • 1789Made this changelog itself a two-tier system, so the public view stays clean and honest while the detail stays private.
• 2026-07-02
  • grenierNow a native Claude Code plugin — run /plugin marketplace add 1789-tech/grenier to install the decluttering skills straight into your agent, no clone needed.
  • grenierAdded a one-line installer (curl … | sh) and a rewritten README with three ways to start: ask your agent, add the marketplace, or run the script.
  • LevelUpThe interactive demo now remembers your place — a page refresh keeps you on the same screen.
  • LevelUpLeave feedback without leaving the demo — lightweight in-app feedback capture.
  • LevelUpSharpened the landing-page positioning — clearer what LevelUp is and who it's for.
  • LevelUpFixed imported practice tracks that were failing to load.
  • 1789Published Terms and Privacy pages; moved GitHub from the top nav into the footer to keep the header focused.
• 2026-07-01
  • grenierNow English-first and multi-country — the skills speak English by default and load a small local recipe (which marketplaces, donation and recycling routes) for your country. France, US and UK today; adding one is a single file.
  • grenierClearer skills — declutter (sort), offload (decide: sell, donate, recycle or toss) and sell (draft the listing), replacing the old sell/listing naming.
  • grenierNo-clone install — paste one instruction into your agent to get started; no git clone or terminal required.
  • grenierLaunched — an open-source collection of Claude Skills for decluttering at grenier.1789.tech; code public and MIT. First skills: declutter-coach, listing-drafter.
  • grenierAdded a no-code quickstart and copy-paste example prompts so non-coders can start decluttering without touching a terminal.
  • LevelUpWaitlist opened — sign up from the landing page to follow the build.
  • 1789Site redesigned — new visual identity, shared nav, portfolio and this Open Startup page.
• 2026-06-30
  • 1789Front door live at 1789.tech — what 1789 is and how it's run.
  • LevelUpLanding page and interactive demo live at levelup.1789.tech.
  • 1789Opened as an open startup — committed to a public changelog, aggregate metrics, and a "what we're building" page, while keeping specs and raw ideas private.