Skip to content
Draft
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion spec/_projects/public-launch/BUSINESS-RULES.md
Original file line number Diff line number Diff line change
Expand Up @@ -7,7 +7,7 @@ epic that implements and proves the rule. Every other epic inherits it.

| | Rule | Owner | Checked where |
|---|---|---|---|
| **PR-1** | When a public package is published, a clean install of it imports under Node with no workspace present, and a package that ships built assets (devtool) ships them | FIX-1635 | FIX-1431's and FIX-1334's acceptance checks, and the closure run FIX-1636 (pack every package, install into an empty project, import each). FIX-1161's scaffold check inherits it: a stranger's first install is the same install |
| **PR-1** | When a public package is published, a clean install of it imports under Node with no workspace present, and a package that ships built assets (devtool) ships them | FIX-1635 | The standing `packed-install` CI job (pack every package, install into an empty project, import each), landed by FIX-1431 and FIX-1334. The closure run FIX-1636 passed on it ([#2543](https://github.com/fixpoint-labs/flow-state-dev/pull/2543)). FIX-1161's scaffold check inherits it: a stranger's first install is the same install |

PR-1 binds two epics because the first-hour scaffold installs from npm, so any package hard gates
leaves unloadable breaks the first hour too. Hard gates owns the fix. The first hour owns only
Expand Down
21 changes: 17 additions & 4 deletions spec/_projects/public-launch/DECISIONS.md
Original file line number Diff line number Diff line change
Expand Up @@ -62,6 +62,18 @@ feature work into this project to make it launch-ready.
- **Fixing a trust-boundary hole never reopens optional org.** Each auth child closes the hole
it names under the principal-owned model (FIX-1442). This binds FIX-1635's children and any
later epic that touches caller identity.
- **The packed-install job proves the release, not the source, and stays.** Settled at FIX-1635's
wrap (FIX-1636, #2543). The standing `packed-install` CI job runs the two-users-one-tenant HTTP
suite against the installed tarballs, permanently. Its workspace-link control must fail. Its
queue leg (leg c) refuses to run without a Redis that answers, rather than skipping. A later
epic that adds an HTTP case adds it to this suite, and never weakens the control or the refusal.
- **BullMQ deployments enforce session concurrency across workers through a Redis lease.**
Settled by FIX-1634. Any epic that puts work into an existing session on a queue host builds on
that lease backend instead of arbitrating concurrency its own way.
- **Hard gates closed without request-id reuse across processes.** That defect is FIX-1665, now
outside the epic as a standalone Public Launch follow-up. FIX-1658 (the lease backend's public
pushed-wake capability) is likewise a standalone follow-up. Neither reopens FIX-1635, and
neither is a hard gate unless Jake adds it (PD-2).

## Open

Expand All @@ -74,13 +86,13 @@ changes are released as patch or minor, and FIX-1192 exists because two changes
bump.

**The trade-off.** 1.0 is the stronger promise to early adopters and reads better in a launch
post. It also makes every public API shape a launch gate, which widens FIX-1635 from known defects
to a full API review before the doors open. With 0.x the launch comes sooner and the gate list
post. It also makes every public API shape a launch gate, which widens the hard-gate bar from known
defects to a full API review before the doors open. With 0.x the launch comes sooner and the gate list
stays as it is, but some teams will wait for 1.0 before they build on it.

![Open fork: 0.x, recommended, beside 1.0. Decides it: FIX-1635 covers known defects under 0.x, and a full API review as well under 1.0. Price of 0.x: some teams wait for 1.0. Flips if someone needs a stability promise on day one](figures/open-launch-version.svg)

It comes down to FIX-1635's scope: 1.0 turns a defect list into a full API review.
It comes down to the hard-gate bar: 1.0 turns a defect list into a full API review.

**My recommendation.** Launch as 0.x, say so plainly in the announcement, and make 1.0 its own
milestone after the first cohort has used the API. Hard gates can then stay a defect list (PD-2).
Expand All @@ -90,4 +102,5 @@ enterprise pilot) that needs a stability promise on day one.

**What being wrong costs.** It's reversible in one direction only. Going 0.x to 1.0 later is
cheap. Launching 1.0 and then breaking an API costs trust with exactly the people we launched to.
It also sets FIX-1635's scope before its objective gate, so it's worth deciding before that gate.
FIX-1635 has wrapped as a defect list, so a 1.0 answer now means a separate API review before
the cut, not a wider hard-gates epic.
14 changes: 8 additions & 6 deletions spec/_projects/public-launch/PLAN.md
Original file line number Diff line number Diff line change
Expand Up @@ -5,12 +5,12 @@
This file covers order, not how anything gets built. It says which order the epics run in, what
each hands the next, and what is deliberately not next. Each epic's own plan owns its checks.

## The arc — as of 2026-09-29
## The arc — as of 2026-10-01

![The arc from August to November 2026. The first hour has an open in-flight bar from Aug 20 to the now line on Sep 29. Hard gates has a lane marked objective in review since Sep 29 and no bar. The docs, brand and demo bar and the launch cut have empty dashed lanes.](figures/arc.svg)
![The arc from August to November 2026. The first hour has an open in-flight bar from Aug 20 to the now line on Oct 1. Hard gates has a short closed bar from Sep 29 to Oct 1, marked wrapped. The docs, brand and demo bar and the launch cut have empty dashed lanes.](figures/arc.svg)

The first hour is the only bar, and it has been open for six weeks. Hard gates' objective went up
for review on Sep 29, and its bar starts when that objective is approved.
Hard gates closed in three days. The first hour is now the only open bar, six weeks in, and it
is what the launch cut waits on.

| Epic | Consumes | Releases |
|---|---|---|
Expand All @@ -20,7 +20,9 @@ for review on Sep 29, and its bar starts when that objective is approved.
| **the launch cut** · *not filed* | All three above, plus the surface framework simplification ships | FIX-1187 the attended first publish, FIX-1192 the changeset repair, the go-live and the announcement |

**Issues in the project with no epic.** The launch cut claims FIX-1187 and FIX-1192. The docs,
brand and demo bar claims FIX-550, FIX-551 and FIX-601. Everything else waits in the
brand and demo bar claims FIX-550, FIX-551 and FIX-601. FIX-1665 (request-id reuse across
processes) left hard gates as a standalone follow-up. FIX-1658 (the lease backend's pushed-wake
capability) is one too, and carries no Linear project. Everything else waits in the
[Linear project](https://linear.app/fixpoint-labs/project/public-launch-2b35a5858733) until an
epic claims it, and none of it is a hard gate unless Jake adds it (PD-2).

Expand All @@ -29,7 +31,7 @@ epic claims it, and none of it is a hard gate unless Jake adds it (PD-2).
- **The launch cut.** It is filed only once hard gates and the first hour are both close to done.
That is also when the launch surface (HN, X, a YC application) and the framework simplification
cut line get asked. Both were open questions in the project's original content, and no epic
needs either answer yet.
needs either answer yet.
- **A second starter template.** One scaffold first. A menu of starters is FIX-548's follow-up if
the single starter turns out to limit adoption.
- **An L2 Workforce child for FIX-1634.** It is filed once the L1 shape lands.
Expand Down
12 changes: 7 additions & 5 deletions spec/_projects/public-launch/SPEC.md
Original file line number Diff line number Diff line change
Expand Up @@ -12,7 +12,7 @@ alone, and not be hurt by a defect we already know about.
|---|---|
| **Winning when** | A stranger with Node installed goes from one command to a working app streaming from a real model, in a new project or one they already have. They never clone our repo or read our source. Every public package installs and imports from npm, and the published surface has no known cross-user leak or silent data loss |
| **The read** | Five checks of "launched", each run from a clean machine. (1) Every public package installs and imports from npm. (2) One command gets you a streaming app. (3) A first flow can be built from the docs alone. (4) No open hard gate. (5) The demo's showcase paths run on the public deploy. Check 5 is about whether they work, not how polished they look (PD-2). None of the five is shown today, because nothing has been published yet |
| **Now** | 0 done · 1 in flight · 1 not started · 2 not filed. The first hour is in development. The hard gates objective went up for review on Sep 29 |
| **Now** | 1 done · 1 in flight · 2 not filed. Hard gates wrapped on Oct 1: its closure run passed against installed tarballs, which is check 4's proof before anything is on npm. The first hour is in development |
| **Kill line** | If the hard-gates list grows faster than it closes, with every pass finding new launch blockers, then the launch definition is too big for one cut. The thing to change is the cut, for example a 0.x release to a small cohort. Relaxing the gates is not the fix |

![The territory. Inside the project: the first hour, the hard gates, the docs, brand and demo bar, and the launch cut. Below the fence is the substrate the launch ships but does not own: framework simplification, which ships first, then framework feature projects, the release pipeline and kitchen-sink. Outside: post-launch growth, speculative platform work, kitchen-sink polish and parked product bets, Ops repair, and the Tier B defect leftovers.](figures/territory.svg)
Expand All @@ -22,12 +22,12 @@ open? Everything below the line ships on its own project's timeline, and the lau
has landed by the cut. Framework simplification is the one exception: the launch waits for it
([Decisions](DECISIONS.md) → decided once).

## The epics — as of 2026-09-29
## The epics — as of 2026-10-01

| Epic | Outcome it owns | State | Surface |
|---|---|---|---|
| [FIX-1161](https://linear.app/fixpoint-labs/issue/FIX-1161) · **first hour** | Someone new gets a working, streaming AI feature from one command in a new or existing project, and their coding assistant writes FSD code that runs | **in flight**: Linear In Development | epic [PR #1301](https://github.com/fixpoint-labs/flow-state-dev/pull/1301) (closed) · branch `epic/building-with-fsd` |
| [FIX-1635](https://linear.app/fixpoint-labs/issue/FIX-1635) · **hard gates** | The known defects that block a public release are closed: trust boundaries, silent data loss, published artifacts that fail to load, lost generator output | **not started**: Linear Backlog. Objective in review | epic spec [PR #2360](https://github.com/fixpoint-labs/flow-state-dev/pull/2360) (open) · GitHub mirror [#2357](https://github.com/fixpoint-labs/flow-state-dev/issues/2357) |
| [FIX-1635](https://linear.app/fixpoint-labs/issue/FIX-1635) · **hard gates** | The known defects that block a public release are closed: trust boundaries, silent data loss, published artifacts that fail to load, lost generator output | **done**: Linear Done Oct 1. Every in-scope child Done; closure QA FIX-1636 passed ([#2543](https://github.com/fixpoint-labs/flow-state-dev/pull/2543)) | epic spec [`specs/epics/FIX-1635/`](https://github.com/fixpoint-labs/flow-state-dev/tree/main/specs/epics/FIX-1635) on `main` · GitHub mirror [#2357](https://github.com/fixpoint-labs/flow-state-dev/issues/2357) |
| FIX-XXX · **docs, brand and demo bar** | A first-time visitor finds one consistent story in the docs, the READMEs and the demo, can build a first flow without help, and finds the showcase paths working | *not filed*. The issues it would claim are in [Plan](PLAN.md) | — |
| FIX-XXX · **the launch cut** | The attended first publish, the go-live, and the announcement | *not filed*. The issues it would claim are in [Plan](PLAN.md) | — |

Expand All @@ -42,12 +42,14 @@ flowchart LR
DB["FIX-XXX · docs, brand and demo bar"] -->|"a site a newcomer can learn from"| CUT
HG -.->|"packages that import · PR-1"| FH
SIMP["Framework simplification · another project"] -.->|"the surface that ships"| CUT
classDef done stroke-width:2px
classDef proposed stroke-dasharray:4 3
class HG done
class DB,CUT proposed
```

Everything flows into the launch cut. Hard gates and the first hour run in parallel, linked only
by a rule, not an order: the scaffold installs the packages hard gates makes loadable
Everything flows into the launch cut. Hard gates has wrapped, so the first hour now builds on
packages already proved loadable, linked by a rule rather than an order
([Rules](BUSINESS-RULES.md) PR-1).

## What this project is not
Expand Down
13 changes: 7 additions & 6 deletions spec/_projects/public-launch/figures/arc.svg
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Loading