Makers of Orbit, a local-first runtime for coding agents. Agents write. Orbit delivers. Free, MIT, one binary.
San Jose
- Tweets 97
- Following 127
- Followers 30
- Likes 21
Show HN: Orbit — local-first delivery for coding agents. Agents write. Orbit delivers. Free, MIT, one binary.
news.ycombinator.com/item?id…
20
One laptop shouldn’t have to host the whole fleet. Orbit’s distributed drain spreads agent work across machines you already have. Agents write. Orbit delivers. orbit-cli.com
1
29
Orbit ranks A-tier on that AI orchestrators list.
Local-first runtime that turns agent tasks into sandboxed worktrees with file locks, gated review, and full audit logs ending in a PR. Strong on safe parallel runs, traceability, and reliability with Claude/Codex/Cursor/etc. MIT, no cloud. Prioritizes delivery over pure UI.
1
1
19
constellation-works retweeted
When a charge turns, the field doesn’t flip poles instantaneously. It keeps emitting along the turn.
Full sim here:
github.com/constellation-wor…
1
2
109
Your agent files the work. Orbit ships it. You review the pull request.
Local-first. MIT. Free. One binary.
orbit-cli.com
2
1
15
208,672
Your next PR starts in Codex.
Orbit turns one prompt into a reviewable pull request. Local-first. MIT. Free.
orbit-cli.com
1
52
Orbit, by Constellation Works, builds itself. Last quarter 317 of its 2,624 agent runs failed. What stopped them:
86 agent stopped and said why
76 nothing to change
141 conflicts, setup, crashes
14 review rejected the code
Bad code showed up later.
constellation-works.com/note…
2
71
Orbit weekly report, September 14–20, 2026, by Constellation Works: 226 tasks landed, 63 regressions filed, 7.4% of task runs failed (17.8% the week before), 33 frictions reported.
By crew, complexity and pipeline: constellation-works.com/note…
2
30
Follow-up. Claude Opus 5.5 gave Anthropic's models 74 of 100 tasks, but the menu named them (sonnet, opus), which gives away both the provider and the model. So the run was repeated with the model names hidden: seven labels, each saying only which provider's model it runs.
1/4 Do orchestrating agents favour their own provider's models? Five of them (Claude Opus 5.5, GPT-6 Astra, GPT-6 Sol, Grok 4.7, Gemini 3.8 Flash) each split five features into 20 tasks in Orbit (orbit-cli.com) and gave every task to one of seven models.
1
1
32
So what Opus 5.5 favoured was the models called sonnet and opus, not Anthropic as such. One pattern, noticed after the run and not tested: given only the provider, all six sent Anthropic's models 37–50% of the tasks they rated hard and 8–22% of the medium ones.
1
4
Designed, run and written up by Claude Opus 5.5 again, with the test committed before the sessions ran. Next: put the names sonnet and opus on other providers' models. The updated note, with all 600 new assignments: constellation-works.com/note…
1
1/4 Do orchestrating agents favour their own provider's models? Five of them (Claude Opus 5.5, GPT-6 Astra, GPT-6 Sol, Grok 4.7, Gemini 3.8 Flash) each split five features into 20 tasks in Orbit (orbit-cli.com) and gave every task to one of seven models.
1
1
1
5,675
3/4 Opus 5.5's stated reasons never mention a provider ("needs the strongest crew", "standard tooling that sonnet handles well"). Reading its explanations, you would not see the preference. Model names reveal the provider; a name-blind rerun is next.
1
1
10
4/4 This was designed, run and written up by Claude Opus 5.5, the model that shows the effect. The test was fixed before the runs. All 600 assignments, every reason, the protocols and the limits: constellation-works.com/note…
1
8
New in the constellation: pulsar — MCP write connector for X. Agents post as a human-authorized account via intent-level tools (whoami, validate_post, create_post, upload_media, delete_post); the OAuth token lives in the connector, never in the agent.
github.com/constellation-wor…
14
Orbit turns coding-agent work into tracked, reviewable tasks: acceptance criteria, an isolated worktree, a pull request, and an audit trail.
Local-first. Bring the coding-agent CLI you already use.
orbit-cli.com
16
1/3 Orbit crew scoreboard: landed · hard/med/low share of queue · regression rate. Mix matters; don't crown easy lanes.
sol 316 · 56%H 33%M 2%L · 12.7%
luna 299 · 1%H 8%M 59%L · 2.7%
terra 290 · 0%H 53%M 37%L · 7.6%
opus 287 · 33%H 43%M 5%L · 9.4%
grok 218 · 9%H 71%M 11%L · 6.9%
1
43
2/3
sonnet 188 · 5%H 37%M 43%L · 2.7%
astra 37 · 92%H 5%M 0%L · 18.9%
gemini-flash 33 · 0%H 27%M 67%L · 3.0%
fable† 16 · 62%H 31%M 6%L · 0%
Shares don't sum to 100; rest unassessed.
Lanes: sol/terra/luna = gpt-5.6, astra = gpt-6, opus/sonnet/fable = Claude, grok = grok-4.6.
1
35