[SandroMaglione]

Summary

• MachineAtom.send/stop and ChildMachineAtom.send/stop no longer drop writes made while the machine is still starting (for example while a bound Atom.runtime layer builds after hydration). Such writes are held, the write result is waiting, and they are delivered in write order once the ref is ready. They settle to Success, to the command failure, or to the startup failure. Later writes queue behind held writes so they cannot overtake them. Writes after stop/completion still fail with StoppedError immediately; child writes still fail with ChildNotActiveError when the owning state is inactive.
• NotReadyError is removed from AtomMachine and from the send/stop failure types. No alias is kept.
• New helper types: AtomMachine.MachineAtom.Any, MachineAtom.State<A>, MachineAtom.Event<A>, ChildMachineAtom.Any, ChildMachineAtom.State<A>. An event-less bridge (Event = never) is assignable to MachineAtom.Any. The internal AnyMachineAtom/FamilyBridge aliases (core and react) now use them.
• Implementation: one ordered command queue per bridge per registry, drained by one fiber that waits with AtomRegistry.getResult. Nothing blocks or busy-waits. The write context gets the registry through an internal keep-alive get.registry atom.
• Tests: new test/unstable/reactivity/AtomMachineCommands.test.ts covers the async-runtime reproduction, write order across readiness, StoppedError after stop, startup failure settling held writes, stop before ready, child writes before parent readiness, and ChildNotActiveError. There is also a React test that uses useAtomSet(machine.send) in a first-mount useEffect with an async runtime, plus typetests for the helpers and command failure types. The new runtime tests fail against the previous implementation.
• Docs: effect-atom-react.md describes early writes and adds a "Generic helpers" section.

Changeset

• Added or updated for a library or package-metadata change
• Not required because this PR does not change a publishable package

One minor changeset for @typeonce/effect-machine. The react package only swaps an internal alias, and its public API is unchanged (same fixed group).

Validation

• pnpm check. Every step passed locally except format:check, because the dprint plugin host is unreachable from the dev sandbox. Changed files were checked with the same dprint plugin versions installed from npm. One unrelated devtools StaticSite test hit its 5s timeout under load and passed when rerun alone.
• Automated type-performance measurement passed or was not required. pnpm perf:types passed locally within budgets.
• Automated runtime- and memory-performance measurement passed or was not required. pnpm perf:runtime was run locally as a smoke test; the CI comparison is authoritative.

---

Generated by Claude Code

[github-actions]

Type performance

Measured with TypeScript 6.0.3 and skipLibCheck=true.

Scenario	Base	PR	Difference
Effect only	55	55	0 (0.0%)
Import effect-machine	55	55	0 (0.0%)
Machine.state (3 children)	2,664	2,664	0 (0.0%)
Machine.make (3 states, 2 events)	8,932	8,932	0 (0.0%)
machine.handle (3 states, 2 transitions)	39,348	39,348	0 (0.0%)
declared transition (10 named branches)	60,165	60,165	0 (0.0%)
atomic construction, guards, and reentry	85,402	85,402	0 (0.0%)
registered invocation (input-taking Effect)	39,467	39,467	0 (0.0%)
machine.handle (depth 24)	194,526	194,526	0 (0.0%)
machine.handle (wide depth 16)	218,233	218,233	0 (0.0%)
machine.handle (parallel/history/choice)	91,633	91,633	0 (0.0%)
machine definition (3 independent implementations)	98,225	98,225	0 (0.0%)
machine exact input/output/error/services	82,791	82,791	0 (0.0%)
execution adapter readiness	71,416	71,416	0 (0.0%)

Marginal instantiations are measured against the matching setup without that API call:

Scenario	Base	PR	Difference
Import effect-machine	0	0	0
Machine.state (3 children)	2,609	2,609	0 (0.0%)
Machine.make (3 states, 2 events)	6,260	6,260	0 (0.0%)
machine.handle (3 states, 2 transitions)	30,416	30,416	0 (0.0%)
declared transition (10 named branches)	51,612	51,612	0 (0.0%)
atomic construction, guards, and reentry	70,062	70,062	0 (0.0%)
registered invocation (input-taking Effect)	32,330	32,330	0 (0.0%)
machine.handle (depth 24)	177,217	177,217	0 (0.0%)
machine.handle (wide depth 16)	201,702	201,702	0 (0.0%)
machine.handle (parallel/history/choice)	79,135	79,135	0 (0.0%)
machine definition (3 independent implementations)	87,544	87,544	0 (0.0%)
machine exact input/output/error/services	70,582	70,582	0 (0.0%)
execution adapter readiness	31,918	31,918	0 (0.0%)

Check times (informational)

Scenario	Base	PR
Effect only	0.02s	0.02s
Import effect-machine	0.02s	0.02s
Machine.state (3 children)	0.05s	0.05s
Machine.make (3 states, 2 events)	0.08s	0.08s
machine.handle (3 states, 2 transitions)	0.22s	0.19s
declared transition (10 named branches)	0.22s	0.24s
atomic construction, guards, and reentry	0.32s	0.30s
registered invocation (input-taking Effect)	0.18s	0.18s
machine.handle (depth 24)	0.35s	0.34s
machine.handle (wide depth 16)	0.43s	0.39s
machine.handle (parallel/history/choice)	0.32s	0.30s
machine definition (3 independent implementations)	0.33s	0.31s
machine exact input/output/error/services	0.31s	0.29s
execution adapter readiness	0.27s	0.26s

Type instantiations are the comparison metric. Check time varies with runner load and is informational only.

[github-actions]

Runtime performance

Median of 5 independent benchmark processes on AMD EPYC 9V74 80-Core Processor with Node v24.22.0.

Pull request baseline

Scenario	Effect Machine
Plan counter transitions	148,861 transitions/s
Drain burst with terminal fence	502,368 increments/s
Drain burst with a change observer	471,768 increments/s
Lookup and send to one child	425,840 increments/s
Start and stop a machine	168,663 machines/s
Start and stop a parent with one child	49,456 families/s
Complete an Effect invocation with tracing enabled	13,071 invocations/s
Complete an Effect invocation with tracing disabled	17,239 invocations/s
Plan transitions through a compound state	131,223 transitions/s
Plan transitions through parallel regions	107,518 transitions/s
Drain burst through a compound state	423,422 events/s
Drain burst through two parallel regions	394,248 events/s
Drain a compound-state burst with a change observer	398,263 events/s

Process runtime reference points

Scenario	Effect Machine
Start and stop a raw generic process	38,257 processes/s
Start and stop a raw compiled process	97,800 processes/s

Memory profile	Effect Machine
Idle machine	1.8 KiB
Raw generic managed process	11.5 KiB
Raw compiled process	3.0 KiB
Two independent idle machines	3.4 KiB
Idle parent with one child	5.9 KiB
Parent with observed child registry	9.4 KiB
Parent with observed invoked child snapshots	6.4 KiB

Effect Machine change from base

Metric	Base	Base variability	PR	PR variability	Difference
Plan counter transitions	149,318 transitions/s	0.5% MAD	148,861 transitions/s	0.6% MAD	-0.3%
Drain burst with terminal fence	506,189 increments/s	0.4% MAD	502,368 increments/s	0.4% MAD	-0.8%
Drain burst with a change observer	472,223 increments/s	0.3% MAD	471,768 increments/s	0.6% MAD	-0.1%
Lookup and send to one child	428,949 increments/s	0.3% MAD	425,840 increments/s	0.1% MAD	-0.7%
Start and stop a machine	169,520 machines/s	1.2% MAD	168,663 machines/s	1.0% MAD	-0.5%
Start and stop a parent with one child	50,279 families/s	1.0% MAD	49,456 families/s	0.0% MAD	-1.6%
Complete an Effect invocation with tracing enabled	13,374 invocations/s	0.6% MAD	13,071 invocations/s	0.2% MAD	-2.3%
Complete an Effect invocation with tracing disabled	17,620 invocations/s	1.5% MAD	17,239 invocations/s	0.7% MAD	-2.2%
Plan transitions through a compound state	132,088 transitions/s	0.2% MAD	131,223 transitions/s	0.9% MAD	-0.7%
Plan transitions through parallel regions	108,110 transitions/s	0.7% MAD	107,518 transitions/s	0.2% MAD	-0.5%
Drain burst through a compound state	424,259 events/s	0.6% MAD	423,422 events/s	1.0% MAD	-0.2%
Drain burst through two parallel regions	395,340 events/s	0.5% MAD	394,248 events/s	0.2% MAD	-0.3%
Drain a compound-state burst with a change observer	397,848 events/s	0.1% MAD	398,263 events/s	0.7% MAD	+0.1%
Idle machine heap per unit	1.8 KiB	4.1% MAD	1.8 KiB	0.3% MAD	+0.2%
Raw generic managed process heap per unit	11.4 KiB	0.1% MAD	11.5 KiB	0.0% MAD	+1.0%
Raw compiled process heap per unit	3.0 KiB	0.1% MAD	3.0 KiB	0.1% MAD	+0.4%
Two independent idle machines heap per unit	3.4 KiB	0.9% MAD	3.4 KiB	0.9% MAD	-0.6%
Idle parent with one child heap per unit	5.9 KiB	0.0% MAD	5.9 KiB	0.0% MAD	0.0%
Parent with observed child registry heap per unit	9.4 KiB	0.0% MAD	9.4 KiB	0.0% MAD	+0.0%
Parent with observed invoked child snapshots heap per unit	6.4 KiB	0.1% MAD	6.4 KiB	0.0% MAD	+0.1%

Process runtime reference change from base

Metric	Base	Base variability	PR	PR variability	Difference
Start and stop a raw generic process	38,242 processes/s	0.3% MAD	38,257 processes/s	0.4% MAD	+0.0%
Start and stop a raw compiled process	98,474 processes/s	1.0% MAD	97,800 processes/s	0.3% MAD	-0.7%

Regression guard

No large, noise-adjusted throughput or heap regressions detected.

Versions and interpretation

• Effect Machine: 0.41.1

Higher throughput is better; lower heap is better. Variability is the median absolute deviation across independent processes, relative to their median. Small differences on shared GitHub-hosted hardware remain informational; the required guard rejects only large changes beyond the measured noise allowance.