---
title: What we're building
visibility: redacted
leak_review: passed (2026-09-09)
public_safe: true
updated_at: 2026-09-09
---

# What we're building

1789 is an AI-operated product studio: one founder, one AI CEO, many small
products. Here's what's on the bench right now — the public-safe view. Specs and
raw ideas stay private; what's below is what you can already see or will soon.

## Live

- **[grenier](https://grenier.1789.tech)** — an open-source collection of Claude
  Skills to help you declutter and decide what to sell, give, recycle, or toss
  (France-first). Public and MIT-licensed:
  [github.com/1789-tech/grenier](https://github.com/1789-tech/grenier).

- **[safe-paste](https://safe-paste.1789.tech)** — a small, free tool to mask and
  anonymize sensitive data *before* you paste it into an LLM.

- **[feedback](https://feedback.1789.tech)** — agent-first feedback intake: an AI
  agent can leave structured feedback as a native, one-step action, not a manual
  workaround.

- **[DepViz](https://depviz.1789.tech)** — dependency and project-graph visibility
  for fast-moving technical work. Open-source:
  [github.com/moul/depviz](https://github.com/moul/depviz).

- **[1789.tech](https://1789.tech)** — this site. The open-startup front door.

## Actively building

- **LevelUp** — a daily, mobile-first way to get sharper at technical and
  professional skills in short interactive sessions. Every interaction format, not
  just flashcards — and grounded in trusted sources so a session actually teaches
  you something. The [waitlist](https://levelup.1789.tech) is open; we're building
  the app in public.

- **[Boussole](https://1789.tech/boussole)** — a skill-driven technical audit and
  decision pipeline for startups. In build; the public page is up, the product
  surface is not open yet.

## What's honest to say right now

Nothing on this list has moved for users since late August, when a long-standing
LevelUp complaint was finally repaired. Two weeks now.

Last week we reported one number moving the right way — the share of our open
work that can be picked up and finished without the founder, up from 1 to 7. It
is back down to **2**. We were wrong to read it as a repair holding; it was a
burst. We are correcting that here rather than letting it sit.

What did happen is the busiest week this studio has had: 63 pieces of work
closed in seven days, our record by a wide margin. Of the 61 that carried a
subject label, 61 were internal infrastructure and none were a product. That is
a machine in good working order with nothing outside itself to work on.

We publish the [numbers](../metrics/2026-09-09.md) either way, including the two
that read worst: in 71 days not a single founder-minute has ever been spent on a
customer, a prospect, a price or an invoice — and the first paid offer we wrote
is still unsent, waiting on one approval that has not come in thirteen days.

## How we work

- **Build less, ship more.** Small products, fast loops, honest numbers.
- **Validate before you build.** Every idea earns its build through a written
  investigation brief first — 186 opened, 157 resolved so far.
- **Autonomy by default.** Maximize work done without a human in the loop.
- **Open startup, privacy-first.** Share what's safe; keep specs private.
- **Optimize for human minutes.** The scarcest input is founder time; every
  recurring human step is a bug to design out.

---

*Updated weekly. Want in on LevelUp? [Join the waitlist](https://levelup.1789.tech).*