7 Sep · 10:24
GitHub · PR #41 · Recorded by source
PR #41 → main
·
Stay clear on what
you’re building.
Artifact keeps track of how what you’re trying to build relates to your decisions, designs, agent work, implementation, and feedback, as those things change.
Mac app. I’ll send you access when it’s ready.
Needs review
1 of 7Team invitations moved ahead of the first preview
Onboarding PRDPrevious version
Let people explore their own project before asking them to invite anyone.
Onboarding PRDUpdated version
Require at least one teammate invitation before opening the project preview.
Onboarding
Motion
00:0006:0012:0018:0024:00
9 Aug24 Aug7 Sep
Through 7 Sep
9 Aug – 7 Sep47 changes
2026-09-07 · 08:30
Founder review notes · Stated by you
A merge alone will not close the customer follow-through question.
·
2026-09-06 · 19:23
Release and pull request record · Recorded by source
PR #40 preserves original occurrences and links later reversals.
·
6 Sep · 16:40
Activation experiment review · Stated by you
18% activation versus 21% control; variant stopped
·
5 Sep · 14:32
Pricing page language · Stated by you
Price unchanged. One plan for the workspace.
·
2026-09-05 · 09:20
Founder review notes · Stated by you
Explain a two-person and a five-person workspace at the same plan price.
·
2026-09-04 · 23:18
Founder review notes · Stated by you
Starting a session does not count as an outcome. Keep activity in Agent Sessions.
·
2026-09-04 · 11:50
Release and pull request record · Recorded by source
PR #39 merged; this does not establish resolution of the separate timeout problem.
·
2026-09-04 · 03:26
Founder review notes · Stated by you
Three of 40 replayed findings retained an outdated evidence revision. The founder recorded the failed check; no correction is established yet.
·
2026-09-03 · 18:52
Release and pull request record · Recorded by source
PR #38 links a collection to its supporting records without generating a new summary.
·
3 Sep · 16:07
First-run flow · Stated by you
A change to the proposed first-run sequence.
·
2026-09-02 · 14:30
Founder review notes · Stated by you
4 of 7 participants incorrectly expected per-seat charges. Keep the price; clarify the language.
·
2026-09-02 · 11:53
Founder review notes · Stated by you
Lead with what changed and what remains unconfirmed, instead of touring every screen.
·
2026-09-02 · 09:17
Founder review notes · Stated by you
5 of 6 participants recovered a failed source connection without assistance; not a production reliability measure.
·
2026-09-01 · 20:13
Release and pull request record · Recorded by source
PR #35 groups repeated reports by underlying issue. Review with real imports still pending.
·
2026-09-01 · 05:42
Founder review notes · Stated by you
A six-hour run imported 18,200 fixture records with two recovered rate-limit retries and one unresolved timeout. The recorded result is not a production reliability claim.
·
2026-08-31 · 10:40
Release and pull request record · Recorded by source
PR #36 merged to distinguish source limits from teammate access.
·
2026-08-29 · 17:26
Founder review notes · Stated by you
Two of five 10,000-event fixtures exceeded the timeout. Incremental retry remains unresolved.
·
2026-08-28 · 14:56
Founder review notes · Stated by you
Do not merge subjects based on name similarity alone; retain the conflicting evidence.
·
2026-08-28 · 10:11
Founder review notes · Stated by you
Three of twenty similarly named subjects were incorrectly combined. Keep uncertain matches separate.
·
2026-08-28 · 00:38
Release and pull request record · Recorded by source
PR #34b adds a retry ceiling after the regression fix. The overnight delivery remains unverified.
·
2026-08-27 · 23:47
Release and pull request record · Recorded by source
PR #34 merged with reconnect-loop tests. A new delivery is not established.
·
2026-08-27 · 15:25
Release and pull request record · Recorded by source
Release 0.8.4 reverted after repeated reconnect prompts. The earlier merge remains part of history.
·
2026-08-27 · 09:10
Release and pull request record · Recorded by source
PR #33 merged to show progress while source records load.
·
2026-08-26 · 16:41
Founder review notes · Stated by you
Four of twelve reviewed findings described the same underlying issue. No precision claim is established.
·
2026-08-25 · 01:07
Release and pull request record · Recorded by source
PR #30 merged after reproducing a lost selection during app restart. This does not establish recovery of every interrupted session.
·
2026-08-24 · 19:36
Founder review notes · Stated by you
Keep monthly billing while willingness to commit annually remains uncertain.
·
2026-08-24 · 14:40
Release and pull request record · Recorded by source
PR #31 merged behind an experiment flag.
·
2026-08-24 · 09:20
Founder review notes · Stated by you
Test a guided setup variant against the existing first-run flow.
·
2026-08-22 · 13:17
Founder review notes · Stated by you
Three of five invitees attended. Two wanted source evidence before trusting the overview; no purchase commitments.
·
2026-08-21 · 16:15
Founder review notes · Stated by you
3 of 8 interviewees preferred annual billing. No purchases or revenue established.
·
2026-08-19 · 22:37
Release and pull request record · Recorded by source
PR #28 stores native IDs so repeated imports do not create new events.
·
2026-08-18 · 14:07
Founder review notes · Stated by you
Do not publish attention items without an inspectable supporting record.
·
2026-08-18 · 02:13
Founder review notes · Stated by you
The scheduled replay produced 6 duplicate events across 2,400 imported records. Recorded in the founder's review of the run; the deduplication fix remains pending.
·
2026-08-17 · 23:18
Release and pull request record · Recorded by source
PR #29 merged to preserve setup progress after a failed connection.
·
2026-08-17 · 18:07
Release and pull request record · Recorded by source
PR #27 merged with clearer plan boundaries; no conversion outcome recorded.
·
2026-08-17 · 09:45
Founder review notes · Stated by you
7 of 12 participants reached a first result, versus 4 of 12 in the earlier flow. Small exploratory sample.
·
2026-08-15 · 12:36
Founder review notes · Stated by you
Organize by product subject instead of equating every repository with a collection.
·
2026-08-14 · 11:20
Founder review notes · Stated by you
Keep one workspace plan; do not charge per invited teammate.
·
2026-08-13 · 11:02
Release and pull request record · Recorded by source
PR #21 collapses long tool output while retaining the original transcript.
·
2026-08-13 · 09:38
Release and pull request record · Recorded by source
PR #20 restores the selected conversation after app restart.
·
2026-08-12 · 16:48
Founder review notes · Stated by you
Test the reconnect-and-review workflow before planning a broader public launch.
·
2026-08-12 · 00:24
Release and pull request record · Recorded by source
PR #19 merged after the late-night import review. Production recovery is not established.
·
2026-08-11 · 22:09
Founder review notes · Stated by you
Remove revoked evidence from published views before the next analysis run.
·
2026-08-11 · 21:43
Release and pull request record · Recorded by source
PR #18 preserves pagination cursors after rate limiting. Backfill completion remains unverified.
·
2026-08-10 · 22:43
Release and pull request record · Recorded by source
PR #24 merged; deployment is not established.
·
2026-08-10 · 10:15
Founder review notes · Stated by you
Let new workspaces explore a sample project before connecting a source.
·
Building got faster.
Judgment matters more.
A note from the builder
AI makes it easier to try an idea and get something working. You can explore several directions, put them in front of people, and learn before committing. But a working version can also make a direction feel settled before you’ve really chosen it.
An agent’s suggestion becomes a feature. An experiment becomes the plan. You keep improving what’s there, while the question of whether it belongs gets harder to see. The product takes shape through decisions you may never have meant to make.
Artifact is being built to keep those decisions connected to the work and the reasons behind it. So you can come back, question what’s taking shape, and decide what to keep, change, or leave out.
Nabil
The work moves on.
The context gets lost.
Codex⌘ K⌄ My project
Refine onboarding flow2h
Continue onboarding work1d
Explore a simpler signup3d
Fix first-run experience5d
Where did I leave off?
Same project. Several Codex sessions. Which one had the thing you need?
− Invite teammates later
+ Invite teammates to continue
Did I decide that?
An optional step is now required. The code changed. You never made that call.
Why did we change this?
The new price is live. The reason is somewhere in an old conversation.
Pricing exploration2 weeks agoLet’s try $19 instead of $12.
What’s behind the change?
We learned that…
On your site$19 / monthLive
Artifact keeps track of how your work fits together
You connect the places where you build and learn: a local repository, GitHub, Codex and other agent histories. Artifact assembles enough context to make that work understandable, and keeps the connections current as things change. Maintaining it is not a job it gives you.
Momentum
See what moved across your tools this month, and the few things you still need to check.
00:0006:0012:0018:00
Product building18GTM6
Signals
Things that connect to each other, across product and GTM, surfaced with the evidence behind them.
Team invitations moved ahead of the first preview
−Let people explore their own project before asking them to invite anyone.
+Require at least one teammate invitation before opening the project preview.
Agent Sessions
Every Codex and Claude session saved in one searchable place, across all your worktrees.
Invitation timeoutPR #41 · MergedLet’s fix the invitation timeout Morgan described. Keep the request running, but show a clear response instead of leaving the page spinning.
I’ll inspect the invitation request and its loading state, then add coverage for a timeout response.
Tool result · test runInvitation request tests: 8 passed.
I’ve opened PR #41 with the response handling and tests. I haven’t verified deployment or retried with the customer.
Collections
The subjects your work is made of, built from the analysis, with everything behind each one.
The first-run flow is being simplified so people can explore their project before inviting teammates. The customer’s result remains unconfirmed.
Source material
First-run flowPRD
Customer conversationTranscript
PR #41 · Invitation timeout fixMerged
Early access
For founders building with AI,
mostly alone or with small teams.
Mac app. I’ll send you access when it’s ready.
Questions I get
Isn’t this Notion, Tana, or a second brain?
Notion stores what you write. Tana reviews AI edits to your notes. Neither knows the difference between what I explored and what I decided. Artifact tracks who said what and what’s committed versus explored, and asks for your attention only when something new changes that.
I already have CLAUDE.md, AGENTS.md and agent recaps. What does this add?
Those carry instructions into a session. They don’t know what changed in Figma, what a customer said, or which PR closed which question. Artifact’s job is the relationship between those things, kept current without you maintaining it.
Does it watch everything on my Mac?
No. You choose the sources: a repository, GitHub, Codex history, the tools you name. Nothing listens, watches, or joins calls on its own.
Where does my data live?
The constraint I’m designing to is local: sources and analysis on your Mac. That’s a constraint to build within, not a guarantee I can make yet.
Does it decide anything for me?
No. It reads, connects and raises. Changing a decision, a document, or a source stays yours to do.
When can I try it?
I don’t have a date I’d put in public. The waitlist gets access first, and prototypes go on the docs site as they exist.