frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

Open in hackernews

Show HN: I benchmarked my memory graph against Memora (0.831 vs. 0.801)

https://github.com/corbym/locomo-recordari
2•colourclash•1h ago
One of the very first problems I hit when I started to do agentic programming was the context problem, and that my agent always started again from the beginning. Being a bit naive and not really knowing what I was doing, I started off writing a mcp tool (because those are cool, right?) to solve the problem.

MD files are primitive, I thought. This memory stuff should really be a database, I thought. So I tried a couple of free memory tools, but nothing was really doing anything for my particular itch. I wanted the agent to remember what I needed, when I needed it, and with the minimal of effort. I wanted it to talk architecture with me, file everything away, and be able to remember it when I wanted to write the stories I needed for the software I was building, and faffing around with plans and files was a pain. I need a knowledge graph.

To be honest, the tools I built are not required if you are happy with an md file solution and your claude.md file. These tools aren't essential for you if that's the case. If you have progressed beyond that set up, carry on reading!

The first tool I wrote was memoryweb (https://github.com/corbym/memoryweb, MIT). In a series of conversations with Claude, we worked out the basics of the api from the agent's perspective and had dogfooded a pretty good tool in just a few weeks. I was using it at work almost constantly. When it went wrong, I asked the agent why, and used memoryweb to record the problem and fix it with a story. It was a very effective loop, and almost effortless.

What about a multi-tenanted version? I had loads of important discussions that my team might like to keep track of. Recordari exists mainly because:

* When two users write conflicting decisions to the same file, no one will notice

* Decisions can be superseded, but old text remains, and the agents can think the stale version is the right version

* After multiple sessions, no one can say why a decision was made or if its true

* Nothing gets recorded about when or who made the decision effectively

Recordari was specifically written to deal with those things. Memories are typed, and the agent can connect memories with typed edges that label nodes "contradicts" and "resolved" - exactly the type of information you'd need to know whether someone disagrees. The server flags semantically close memories that could conflict, agents then adjudicate the possible contradictions and resolves them. Every memory has an owner, so agents don't just overwrite human decisions.

Oh right, the numbers. Yeah.. so the numbers were run using Memora's (Microsoft Research, ICML 2026) open-source benchmark harness, unmodified. The same model, the same LoCoMo dataset and the same category-exclusion conventions. Recordari scored 0.831 end to end, the LLM judge accuracy vs Memora's published 0.801. All the results are public (MIT): https://github.com/corbym/locomo-recordari. The biggest surprise was the multi-session questions coming out at 0.869 vs the raw storage baseline of 0.760.

The harness itself runs against the prod API - exactly the MCP search and recall api that a real agent would use. Not a mocked backend, the real API, with a configuration of `top_k=30`, and a 1 hop edge expansion.

Recordari works with Claude, Claude Code, ChatGPT and pretty much any agent that can connect to MCP and reason.

If you want to run an agent against it to run the LoCoMo benchmark yourself, you can anon login at:

https://admin.recordar.io/start

Just remember to add the Full Skill from the Connect page, and then Connect your agent. Et Viola.

What I would like from HN: have you felt the same pain with a team as me? I'd also like to know if this isn't something you'd want.

Thanks everyone!

Matt

Comments

colourclash•1h ago
--

References and Further reading:

* https://github.com/corbym/memoryweb (MIT)

* Recordari commercial site: https://recordar.io/

* Recordari Admin (register/sign in page): https://admin.recordar.io/

* Recordari Anon login (team tier): https://admin.recordar.io/start

* Microsoft blog: https://www.microsoft.com/en-us/research/blog/memora-a-harmo...

* Memora github: https://github.com/microsoft/Memora

* Memora paper: https://arxiv.org/abs/2602.03315

Things to ask the agent when running the demo sandbox:

* Why was VTIR created?

* Describe what was next in the project and why?

* Why must Pascal fixtures be written before the Rust port?

* What is TurboSound and why does it need special handling?

* What is the current state of the project?

* Why does WASM file I/O work differently from native?

* What are the standing rules for the AY chip port?

Song to download and try on VTIR (https://corbym.github.io/vtir/ - live web demo)

* https://zxart.ee/tune/325740 (download pt3 original file)

colourclash•12m ago
Also I'm here and happy to answer anyone's questions about the methodology.

AI Startup Cognition in New Funding Talks at $40B Value

https://www.bloomberg.com/news/articles/2026-08-12/ai-startup-cognition-in-new-funding-talks-at-4...
2•closetheloopdev•1m ago•0 comments

Qwen3.8-27B Countdown

https://modelscope.cn/models/Qwen/Qwen3.8-27B
1•apitman•1m ago•0 comments

The Odyssey as AI Allegory

https://unherd.com/2026/08/the-odyssey-as-an-ai-allegory/
1•voxleone•2m ago•0 comments

Show HN: Kery – comments on your PR with a video of the feature working

https://github.com/Kery-HQ/Kery
1•Kvlshah•3m ago•0 comments

Is GitHub Online?

https://isgithub.online/
2•vednig•4m ago•1 comments

Canadians call for U.S. Ambassador Hoekstra to be expelled

https://www.cbc.ca/lite/story/9.7303372
4•colinprince•5m ago•0 comments

Show HN: Capto – alerts when a background job goes quiet, with breadcrumbs

https://capto.run
1•johndory80•5m ago•1 comments

We made Kamal deploy to Kubernetes without touching deploy.yml

https://miget.com/blog/deploy-rails-with-kamal
1•ktaraszk•5m ago•0 comments

Hax – a minimalist, terminal-native coding agent written in C

https://usehax.dev/
1•OleksandrC•5m ago•0 comments

License Plate Reader Searches Should Require a Warrant

https://andrewpwheeler.com/2026/08/12/license-plate-reader-searches-should-require-a-warrant/
7•apwheele•5m ago•0 comments

StashKit – ez QR code generator

https://stashkit.co
1•adStein•6m ago•0 comments

I built an AI website in 3 hours, got 562 sign-ups, and felt lost

https://medium.com/@kakuxwn/i-built-an-ai-website-in-3-hours-it-got-562-sign-ups-in-8-hours-and-l...
1•Irving-AI•8m ago•1 comments

Aperiodic Tilings

https://presheaf.blogspot.com/2021/08/aperiodic-tilings-some1-explainer.html
1•unprovable•8m ago•0 comments

Where Did the Productivity Gains Go?

https://idiallo.com/blog/where-did-the-productivity-gains-go
3•Brajeshwar•9m ago•0 comments

Show HN: TokenMaxxer – Create AI spend leaderboards with your friends

https://tokenmaxxer.xyz
2•SYeomans•11m ago•0 comments

Water-from-Air Systems to Enable Decentralized Water Infrastructure

https://www.kubota.com/news/2026/20260722-001220.html
1•beauzero•11m ago•0 comments

Every disease is a policy failure

https://worksinprogress.co/issue/future-of-medicine/
3•bensouthwood•11m ago•0 comments

I ran the same local LLM on an RTX 5070 laptop and one with an iGPU

https://www.xda-developers.com/same-local-llm-on-rtx-5070-and-integrated-graphics/
1•vednig•12m ago•0 comments

US Treasury ends ownership reporting rules for U.S. companies

https://www.washingtonpost.com/business/2026/08/11/treasury-ends-ownership-reporting-rules-us-com...
2•toomuchtodo•13m ago•1 comments

I wrote an AI textbook – how long until AI can do it better?

https://www.interconnects.ai/p/i-wrote-an-ai-textbook-how-long-until
1•kurinikku•15m ago•0 comments

Simulacra and Simulation

https://en.wikipedia.org/wiki/Simulacra_and_Simulation
3•soupspaces•15m ago•1 comments

Graduate Student Proves the Fractal Uncertainty Principle

https://www.quantamagazine.org/graduate-student-proves-the-fractal-uncertainty-principle-20260812/
1•Tomte•16m ago•0 comments

iDrive Ads

https://tumbledry.org/2026/08/07/idrive_ads
1•JumpCrisscross•16m ago•0 comments

Distrosea – Test drive Linux distros online

https://distrosea.com/
2•robtherobber•17m ago•0 comments

Ultrarare Nintendo Mario game cartridges discovered in Wisconsin

https://apnews.com/article/nintendo-wisconsin-mario-cartridges-video-de88a9fd5643913454328830e73f...
2•andsoitis•18m ago•0 comments

SVGcad

https://www.svgcad.com/
1•quaestio•18m ago•0 comments

Give Your Baby a Machete

https://slate.com/human-interest/2013/04/bad-parenting-ideas-that-are-actually-good-for-some-babi...
2•tdeck•18m ago•0 comments

Evaluating Your Agentic Harnesses

https://data4sci.com/blog/evaluating-your-agentic-harnesses
1•Anon84•19m ago•0 comments

ZZBoard: Open-source "shared work" network for autonomous agents.

https://www.zzboard.com/
1•homanp•19m ago•1 comments

A Built AI-Powered MVP – Available as a One-Time Opportunity

https://www.indiehackers.com/post/a-fully-built-ai-powered-mvp-available-as-a-one-time-opportunit...
1•AutFlowStudio•21m ago•0 comments