Elves - They work while you sleep

Elves v1.10.1 is out: Elves Reports! When the workers finish an overnight run, they now leave the manager a proper morning report in HTML format, not just a PR and a raw execution log. Reports include status, problems found, lessons learned, validation proof, residual risks, and collapsible batch ledgers. They work while you sleep. Now they leave the report on the desk. github.com/aigorahub/elve…

Elves v1.9.0 is live! This release is about cleaner long runs: final cumulative review before handoff, strategic forgetting, memory/resource hygiene, and safer Codex/Claude cleanup guidance. Using Codex Goals? Stage Elves normally, then launch the prepared prompt with /goal. Goals keeps Codex moving. Elves defines “done”: Stop Gate, PR review loop, memory hygiene, and the final Readiness Gate.

Elves v1.8.0 is out! This release hardens long unattended runs: - explicit Stop Gate + continuation state - warning-only survival guide validation - clearer checkpoint semantics - stronger “don’t stop / don’t coast” guidance for Claude Code + Codex

v1.7.0: The Elves Remember Until now, every elf run started from scratch. The elves knew the plan but not what they learned last time. Now they do. Learnings persist across sessions. The codebase improves over time. Durable docs describe your codebase for future elves. The tenth elf run is now dramatically better than the first. Plus: stronger regression discipline, installation management, and docs-in-the-loop review gates. Read more in the latest article by @johnennis:

An update to @elves_skill to improve the memory management from session to session!

Elves just got a reliability upgrade! The big change is a new stage first, launch second workflow for unattended runs. If a user pastes a giant plan and says “run now,” Elves now slows down, stages the run, cleans the docs, lines up the branch and PR, runs preflight, and waits for a final launch command. Then the real overnight run starts from a short hard prompt with momentum. This fixes one of the most common failure modes: giving the model too much at once and having it lose the thread. What shipped: - v1.6.0 introduced the new stage then launch flow - v1.6.1 cleaned up review feedback and tightened internal consistency Both the Claude Code and Codex versions were updated, along with the README, kickoff templates, survival guide template, and execution log template. Release notes: github.com/aigorahub/elve…

Let us know if you would like this feature added!

This is awesome, but the idea that it’s going to turn Claude Code into a 24/7 programmer is wrong The model is still going to hit context drift, and forget what it is doing That’s why you need @elves_skill

Testing @johnennis Elves tonight. Let‘s see what these wizards put on the table by tomorrow morning :)

Been using @elves_skill by @johnennis lately and the "Ralph Loop" workflow is the real deal. Instead of fighting for the perfect one-shot prompt, you just let the agents iterate, fail, and self-correct until the code is actually solid.

I don't think I'll ever get used to working this way Similar to FSD, it's just so magical

@hamzadiazbtc @johnennis Thank you!

This weekend, running @elves_skill on 7 projects across 3 machines, I easily got at least a year’s worth of work done It might have been more like two or three years had I done it all myself Things are about to go absolutely crazy as more and more people start to work like this

Elves v1.3-1.4: Here Comes the Judge. Your code review tool checks whether the code is correct. It doesn't check whether the app still does what it's supposed to do. Now Elves does. Write a constitution. The judge enforces it on every batch. New article by @johnennis github.com/aigorahub/elves

Hello world, we are here to help!

@johnennis @GitKraken GitKraken and the Elves go great together!

I was using @GitKraken today to review last night's Elves runs, and I was reminded again just how great a tool it is If you are agentic coding, you absolutely should be using GitKraken to keep up to speed on what's going on No other tool on the market comes close