Show HN: Remoroo — trying to fix memory in long-running coding agents

What it claims

A Show HN post has introduced Remoroo, a tool that, it has been reported, runs autonomous research on your code locally overnight. Point it at a spec file (program.md), and Remoroo allegedly runs experiments, edits, tests, evaluates, keeps or reverts changes. Wake up to better results, the pitch promises — sweet dreams for busy devs. Who wouldn't want to hand off the tedious tinkering and wake up to improvements?

How it works (allegedly)

Remoroo asks for a spec and then runs through variations against your repository while you sleep. The startup frames this as an attempt to address the classic problem of "memory" in long-running coding agents — the drift and fragility that plagues autonomous systems like the various Auto-GPT forks and agent experiments we keep seeing. It has been reported that the system runs locally on your machine and emphasizes safe rollbacks, but those operational details will invite scrutiny from teams that worry about overnight writes to source code.

Pricing and reception

The Remoroo site lists a free tier that includes monthly run credits; paid plans are also shown for heavier usage. The Show HN thread is drawing attention from developers curious about whether this approach actually improves agent reliability or just automates late-night fiddling. Early reactions are mixed: some see the promise of automated, experimental nightly runs; others raise a hand and ask, "Who’s reviewing these edits before they land?"

In short: clever idea, familiar caveats. Automating experiments overnight could be a genuine productivity win — if trust, safety, and reproducibility are nailed down. Sleep on it, literally — but maybe don't let it loose on your critical branches until you've kicked the tires.

Sources: remoroo.com, Hacker News