frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

fp.

Open in hackernews

Show HN: Fava Trails – Git-backed memory for AI agents using Jujutsu (JJ)

https://github.com/MachineWisdomAI/fava-trails
3•timeleft--•2h ago
Hey HN,

I've been building and running autonomous AI agents (recently consulting on systems testing high on MLE-Bench), and I kept hitting the exact same architectural wall: memory poisoning.

Right now, the industry standard for agent memory is to dump text into storage with very little thought about correctness. If an agent hits a transient network error and writes "this environment has no GPU" to its memory, and later realizes it actually does have a GPU and writes a correction... a standard vector search returns both. Your agent is now schizophrenic, holding contradictory beliefs because they are semantically similar.

Furthermore, memory systems are mostly "write-through." If a user jailbreaks your bot into an offensive persona, the bot saves that as a "user preference" and it persists across sessions forever. We don't let untested code reach main, so why do we let unvalidated agent thoughts reach shared memory?

I built FAVA Trails to fix this. It's an agentic memory layer that uses Jujutsu (JJ) version control under the hood.

For the version control nerds: why Jujutsu? I originally looked at standard Git and SQL-based VCS like Dolt. But JJ is the perfect substrate for autonomous agents. Its conflict resolution, first-class operation log, and the fact that the working copy is a commit makes it inherently crash-proof for long-running agent scripts. If an agent session crashes mid-thought, the JJ commit is already there. No detached HEAD nightmares, no staging area rituals for the agent to mess up. Just atomic state snapshots. (It's colocated with Git, so you can still push the data to a standard remote).

How it works:

- Draft Isolation: Agents write to a local draft namespace first. It doesn't pollute shared memory.

- Trust Gate: A mandatory promotion workflow. An independent LLM (or explicit human approval) reviews the draft before it merges into canonical truth.

- Supersession Chains: Corrections don't silently overwrite history. They link back to it, so you get a full causal graph of why the agent changed its mind.

- MCP Native: It runs as a Model Context Protocol (MCP) server, so agents interact with it via semantic tools (recall, save_thought, propose_truth) and never run VCS commands directly.

It's Apache 2.0 and strictly a pip-installable tool (no cloud lock-in, the data is just Markdown files with YAML frontmatter in your own repo).

Repo: https://github.com/MachineWisdomAI/fava-trails Case Study/Docs: https://fava-trails.org

I'd love to hear your thoughts on using JJ as a backend for state, or how you're handling the "gaslighting agent" problem in your own multi-agent stacks.

How I'm Using Local Large Language Models

https://www.jvt.me/posts/2026/02/28/local-ai/
1•zdw•4m ago•0 comments

Qwen3.5-35B-A3B-GGUF from Unsloth

https://huggingface.co/unsloth/Qwen3.5-35B-A3B-GGUF
1•vincirufus•4m ago•0 comments

Monitor the Situation

https://monitor-the-situation.com/middle-east
1•bookofjoe•4m ago•0 comments

Claw21 – Blackjack for Agents

https://clawhub.ai/stainlu/blackjack
1•stainlu•6m ago•1 comments

C-Mera

https://github.com/kiselgra/c-mera
1•tosh•8m ago•0 comments

Attention Military Members and Veterans: You Have Been Defrauded

https://beendefrauded.substack.com/p/attention-military-members-and-veterans
1•htwatchdogs•12m ago•0 comments

The March of Nines

https://sruthipoddutur.substack.com/p/the-march-of-nines-part-1
9•sruthipsr•12m ago•0 comments

Kbtz: A task tracker and TUI workspace for coding agents

https://github.com/virgil-king/kbtz
1•ostrich_network•13m ago•0 comments

Switch between different Claude Code profiles

https://github.com/kimrgrey/claudini
1•kimrgrey•14m ago•0 comments

Wish List: SSH Keys in Passwords

https://sixcolors.com/post/2026/01/wish-list-ssh-keys-in-passwords/
1•watermelon0•15m ago•0 comments

Tablesaw: Java Dataframe and Visualization Library

https://github.com/jtablesaw/tablesaw
1•saikatsg•15m ago•0 comments

Who Will Command the Robot Armies? (2016)

https://idlewords.com/talks/robot_armies.htm
1•diafygi•15m ago•1 comments

Show HN: I dump all my private notes into an LLM and tell it to build me a site

https://tresbuchet.com/
1•elijahlucian•15m ago•0 comments

If code is cheap, intent is the currency

https://zknill.io/posts/commit-message-intent/
1•zknill•17m ago•0 comments

In the Northwest, Polyamory Finds Something New: Legal Protection

https://www.nytimes.com/2026/02/28/us/pacific-northwest-polyamory-legal-protection.html
1•geox•19m ago•0 comments

Ask HN: If AI boosts productivity so much, why would a company lay anyone off?

1•ex-aws-dude•20m ago•2 comments

Jargonism: A Business English Dictionary

https://jargonism.com
1•hliyan•23m ago•0 comments

AI writing just isn't good enough – and if you're using it, everyone can tell

https://www.theglobeandmail.com/opinion/article-ai-writing-just-isnt-good-enough-and-if-youre-usi...
2•paulpauper•23m ago•0 comments

Secretary of Defense bans military from attending Ivy League Schools

https://www.foxnews.com/politics/hegseth-bans-military-attending-princeton-columbia-other-elite-u...
2•hmokiguess•24m ago•0 comments

Show HN: Mdspec.dev – open-source spec management platform for technical teams

https://mdspec.dev
1•zameermfm•24m ago•0 comments

If A.I. Is a Weapon, Who Should Control It?

https://www.nytimes.com/2026/02/28/opinion/artificial-intelligence-trump-hegseth-anthropic.html
2•paulpauper•24m ago•0 comments

What Your DNA Reveals About the Sex Life of Neanderthals

https://www.nytimes.com/2026/02/26/science/human-evolution-neanderthal-sex.html
1•paulpauper•25m ago•0 comments

Show HN: Which VCs are Tier 1?

https://vc-compare.vercel.app/
1•nezaj•27m ago•0 comments

Last Year in Container Security

https://nanovms.com/blog/last-year-in-container-security
2•valyala•27m ago•0 comments

Should AI chatbots have ads? Anthropic says no

https://arstechnica.com/ai/2026/02/should-ai-chatbots-have-ads-anthropic-says-no/
1•joozio•27m ago•0 comments

Agentic Engineering – Choosing the Right Level of Guidance

https://potocki.dev/posts/agentic-engineering-choosing-the-right-level-of-guidance/
1•tazer•28m ago•0 comments

GPL as the best licence

https://blog.hansenpartnership.com/gpl-as-the-best-licence-governance-and-philosophy/
2•lr0•28m ago•0 comments

Show HN: AIQuotaBar – See Claude/ChatGPT usage limits in your macOS menu bar

https://github.com/yagcioglutoprak/AIQuotaBar
2•toprak123•30m ago•1 comments

FreeBSD Home NAS, Part 14: Logs with VictoriaLogs and Alerts with VMAlert

https://rtfm.co.ua/en/freebsd-home-nas-part-14-logs-with-victorialogs-and-alerts-with-vmalert/
1•setevoy•30m ago•0 comments

Increase of AI bots on the Internet sparks arms race

https://www.wired.com/story/ai-bots-are-now-a-signifigant-source-of-web-traffic/
1•joozio•30m ago•0 comments