frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

fp.

Open in hackernews

Three Memory Architectures for AI Companions: Pgvector, Scratchpad, Filesystem

https://www.emotionmachine.com/blog/how-memory-works
2•sarbak•2h ago

Comments

gkanellopoulos•1h ago
Your gradual transition from v1 to v3 is a common pattern I've seen elsewhere. Project teams usually start with retrieval then face recall quality problems at which point they start wondering whether or not to let the LLM and its context window take over. This is a natural and instinctive approach but imho there are two issues with that approach. Firstly the LLM decides what matters and that decision is irreversible and secondly it does not scale well over time. A couple of months after when the user will ask "What did I ask you to remind me about X 3 months ago?" the summary might have rotated by then that detail out.

I agree that there is a fundamental issue with v1-style retrieval but in my view is not the scoring formula but the fact that similarity search mixes semantically related with really valuable data. For example a memory about "surfing last weekend" and a memory about "wanting to surf one day in Hawaii" will both score high in the question "What outdoor activities do I like?". However, in the question "what did I do last weekend?" only one is useful while both will appear in the injected context. One way that might solve this issue is by introducing more retrieval dimensions like keyword-matching (BM25), entity-aware scoring and temporal signals in order to then determine which memories are truly relevant to the users question. This of course adds up during ingestion but in general, async ingestion is underrated. Generally speaking users expect near-instant responses while ingestion can be slower.

If I may ask, have you done any benchmarking on the v3 approach? It would be interesting to see how a v3-style solution handles factual vs general preference questions. This usually is a tricky one for memory systems.

Vibe Analysis

https://newsletter.dancohen.org/archive/vibe-analysis/
1•apollinaire•1m ago•0 comments

I built a tool to control coding agent on my Mac through my phone

https://www.tarsy.dev/
1•leddo•1m ago•0 comments

Miracle – A hackable window manager for Linux

https://github.com/miracle-wm-org/miracle-wm
1•matthewkosarek•2m ago•0 comments

Iran raises suspicion of 'uranium theft' plot after clashes with US forces

https://www.aa.com.tr/en/middle-east/iran-raises-suspicion-of-uranium-theft-plot-after-clashes-wi...
1•bhouston•3m ago•0 comments

SubCat: Stop babysitting your Pull Requests

https://subcat.todaywedream.com/
1•semisse•4m ago•1 comments

NASA's Artemis II Lunar Crater Recognition Flash Cards [pdf]

https://ares.jsc.nasa.gov/course-content/lunar_flashcards/CommonLTPTargets_Flashcards.pdf
1•reaperducer•4m ago•0 comments

Outlook for Public and Private Markets

https://www.apollo.com/wealth/the-daily-spark
1•toomuchtodo•4m ago•1 comments

Show HN: u there? Let Claude Code check if you're at your computer

https://github.com/odfalik/uthere
3•odedfalik•5m ago•0 comments

Stages of Denial

https://nsl.com/papers/denial.html
1•tosh•5m ago•0 comments

US Rescue op in Iran was likely attempt to extract HEU from Isfahan

https://twitter.com/TheNavroopSingh/status/2041081690769305729
1•bhouston•5m ago•1 comments

Processing 3GiB of JSON in the browser in 2 seconds via WebAssembly

https://refi64.dev/posts/discordsona.html
2•tosh•6m ago•0 comments

10 Claude Code Principles

https://jdforsythe.github.io/10-principles/
2•topherjaynes•6m ago•0 comments

Complete biosynthesis of psychedelic tryptamines from three kingdoms in plants

https://www.science.org/doi/10.1126/sciadv.aeb3034
1•Stratoscope•7m ago•0 comments

A Drop in for `anyhow:Context` that reports the failure location

https://gist.github.com/alpaylan/ddac04fe80219fb2f1ba5c46dd5647af
1•alpaylan•8m ago•0 comments

Ask HN: First Step to Utopia – how can we turn normies into AI power users?

1•dl92•8m ago•0 comments

Provider-Dependent Energy Effects of Prompt Compression

https://arxiv.org/abs/2603.23528
1•PaulHoule•12m ago•0 comments

Show HN: Yume, a Desktop GUI for Claude Code (Tauri/React)

https://aofp.github.io/yume/
1•yuruko•12m ago•0 comments

Join HN: The Kernel Dev Community in SF

https://www.thekernel.dev/
1•Exorust•13m ago•0 comments

Agent Reading Test

https://agentreadingtest.com
1•kaycebasques•14m ago•1 comments

BatteryNotifier – Cross-platform app with smart notification escalation

https://github.com/sandip124/batterynotifier
1•sandip124•16m ago•0 comments

Absurd: Simplest durable execution workflow system

https://github.com/earendil-works/absurd
3•nkko•16m ago•0 comments

OpenRockets Foundation Released Magazine V2

https://mag.openrockets.com
2•openrockets•16m ago•0 comments

SOM: A minimal Smalltalk for teaching of and research on Virtual Machines

http://som-st.github.io/
1•tosh•17m ago•0 comments

The Big Bang: A.I. Has Created a Code Overload

https://www.nytimes.com/2026/04/06/technology/ai-code-overload.html
1•donohoe•17m ago•0 comments

Zooming UIs in 2026: Prezi, impress.js, and why I built something different

5•tinchox6•18m ago•0 comments

Why some cities are shutting down Flock cameras amid privacy concerns

https://www.theguardian.com/us-news/ng-interactive/2026/apr/06/flock-cameras-privacy-concerns
6•devonnull•20m ago•0 comments

Tell HN: Reddit Seems to Be Down

5•forlorn•21m ago•2 comments

The Edwardian builder who had changed how we keep time

https://www.bbc.com/news/articles/cd9vxg95ze2o
1•speckx•22m ago•0 comments

Scientists identify 'neural fingerprint' of psychedelic drugs in the brain

https://www.theguardian.com/science/2026/apr/06/scientists-identify-neural-fingerprint-of-psyched...
2•tmnvix•22m ago•0 comments

Refactoring Untested Code

https://yusufaytas.com/refactoring-untested-code/
10•kakerane•22m ago•0 comments