frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

KOReader

https://koreader.rocks/
276•Cider9986•3h ago•97 comments

Handbook.md shows that long policy documents do not reliably govern agents

https://arxiv.org/abs/2607.25398
57•spIrr•1h ago•32 comments

Show HN: Bullshit Detector – agent skills that fact-check videos and articles

https://github.com/SerhiiKorniienko/bullshit-detector
34•skorniienko•1h ago•14 comments

Document-borne AI worms can self-propagate through Copilot for Word

https://enklypesalt.com/posts/context-collapse-part3-ai-worming-through-word/
139•Canopy9560•2h ago•117 comments

More Tailscale tricks for your jailbroken Kindle

https://tailscale.com/blog/jailbroken-kindle-proxy-tun-modes
296•Error6571•9h ago•95 comments

Disrupting supply chain attacks on NPM and GitHub Actions

https://github.blog/security/supply-chain-security/disrupting-supply-chain-attacks-on-npm-and-git...
23•nyku•2h ago•7 comments

Amiga Graphics Archive

https://amiga.lychesis.net/index.html
74•Bluestein•3h ago•13 comments

A Texture Lookup Approach to Bézier Curve Evaluation on the GPU (JCGT)

https://jcgt.org/published/0015/02/01/
16•ibobev•1h ago•2 comments

SpecForge – A Platform for Authoring Formal Specifications

https://docs.imiron.io/v/0.5.10/en/tour.html
45•agnishom•3h ago•2 comments

User Interfaces of the Demo Scene

https://www.datagubbe.se/scenegui/
295•zdw•9h ago•51 comments

SQLite in Production: Optimizing WAL Mode, Concurrency, and VFS Layers

https://micrologics.org/blog/sqlite-in-production-optimizing-wal-mode-concurrency-and-vfs-layers-...
143•ankitg12•6h ago•52 comments

ALP: Adaptive lossless floating-point compression

https://github.com/cwida/ALP
17•fanf2•3d ago•0 comments

Lisp moving Forth moving Lisp

https://letoverlambda.com/textmode.cl/guest/chap8.html
72•fallat•2d ago•20 comments

Codex Security

https://github.com/openai/codex-security
549•bakigul•17h ago•199 comments

French musician Kavinsky found dead

https://www.euronews.com/culture/2026/07/29/dj-kavinsky-known-for-his-track-nightcall-found-dead-...
149•bristleworm•2h ago•36 comments

Show HN: Echologue – the private AI voice journal I built for myself

https://echologue.com/
13•arisAlexis•1h ago•8 comments

Hunter-gatherers introduced fish to a mountain lake 7000 years ago

https://www.newscientist.com/article/2580119-hunter-gatherers-introduced-fish-to-a-mountain-lake-...
8•stevenwoo•2d ago•4 comments

Darktable

https://www.darktable.org/
7•siatko•1h ago•0 comments

Ancient Rome's version of Google Maps: how long to reach the beach

https://www.euronews.com/culture/2026/07/02/ancient-romes-version-of-google-maps-how-long-to-reac...
44•gnabgib•6d ago•32 comments

Show HN: I was tired of opening 2 tabs for every HN link, so I made a userscript

https://github.com/twalichiewicz/HNewhere
374•twalichiewicz•16h ago•102 comments

Substack writers, you need a website

https://elizabethtai.com/2026/06/10/substack-writers-you-need-a-website/
601•speckx•21h ago•315 comments

Shipping Godot VR and Porting to PSVR2: A Partial Post Mortem

https://www.claire-blackshaw.com/blog/2026/07/shipping-godot-vr-and-porting-to-psvr2-a-partial-po...
7•ibobev•1h ago•0 comments

Half-Life ported to Mac OS 9

https://mac-classic.com/news/half-life-ported-to-mac-os-9/
284•freediver•17h ago•136 comments

The only road in London where you have to drive on the right

https://www.mylondon.news/news/transport/only-road-london-you-drive-30499294
19•theanonymousone•4h ago•25 comments

Show HN: Vimgolf.ai – Learn Vim by playing through a map of levels

https://vimgolf.ai
27•nickandbro•4h ago•18 comments

Cracking Windows Open: Porting RADV to Win32

https://www.collabora.com/news-and-blog/news-and-events/cracking-windows-open-porting-radv-to-win...
76•zdw•9h ago•21 comments

ReFrame – The EPaper Camera

https://reframe.camera/
177•phil294•14h ago•40 comments

Steel Bank Common Lisp version 2.6.7

https://sbcl.org/all-news.html?2.6.7
255•tmtvl•21h ago•121 comments

LearnVector – Andrew Ng's AI company building one‑to‑one learning experiences

https://learnvector.ai/
236•ajhai•12h ago•151 comments

Delayed Gratification – Proud to Be 'Last to Breaking News'

https://www.slow-journalism.com/
321•speerer•22h ago•186 comments
Open in hackernews

Handbook.md shows that long policy documents do not reliably govern agents

https://arxiv.org/abs/2607.25398
54•spIrr•1h ago

Comments

leetrout•42m ago
HANDBOOK.md is a benchmark for long-context agentic instruction following, modeled on how enterprise employees follow company handbooks in their day-to-day work. Each task is a unique RL environment with internal tools and external MCP servers, spanning five enterprise domains: Finance, Medical Billing, Insurance, Logistics, and HR.

The prompts reflect the actual jobs enterprise workers perform every day. Each task drops an AI agent into a live company environment, requiring them to cross-reference an extensive, multi-section handbook against a cluttered inbox, a multi-channel Slack workspace, Jira queues, and a stack of files (spreadsheets, PDFs), and working out both what to do and what the handbook forbids.

https://github.com/surge-ai/handbook/tree/main

mcdeltat•41m ago
Yeah checks out with my anecdotal experience with Claude. It is pretty great at following instructions - for about 10 minutes, after which it seems to ignore things I told it before.

I have quite explicit and strong instructions (e.g. don't write massive comments, use existing functionality, etc.) in CLAUDE.md files which seem to get bypassed surprisingly quickly when doing real tasks. Yet if I tell it these things in a prompt during the task, it performs way better.

Result is I'm trying to resist adding more and more things to CLAUDE.md files which in some scenarios it does well but in other scenarios totally ignores and messes up.

cyanydeez•32m ago
I believe the correct static instructions are about getting it at the right starting point for whatever class of projects you're working on; not as a continued referencable or "HOWTO" of what it's doing. They're all just "grooming" the LLM for future instructions.

The coding harness is what's getting it to continually align to your current instructions.

This is very obvious with local models.

spIrr•29m ago
As a hobbyist, I find it difficult to figure out how to make Claude stick with some repeating things I want it to do after every major action, like re-evaluate the completeness of tests, update the documentation, etc. And CLAUDE.md/AGENTS.md definitely did NOT help there, sadly.
Muromec•12m ago
Don't use the default harness, write your own instead.
Supermancho•7m ago
This is the way. Making your own agent to have a sticky memory context that is prepended to every execution is necessary to ensure each task is bounded by those precepts.
victorbjorklund•7m ago
Hooks can be pretty useful for that. A hook when it is finished ”run tests suite and check coverage” ”check if your changes require updating the docs”
mcdeltat•
firasd•40m ago
Opus 4.8 (max thinking) scored highest and Grok 4.3 lowest

It's hard to understand what's going on with Grok. It's like it has capabilities in a theoretical sense but maybe the training is so focused on being in x.com/grok.com with the web search tool enabled for "is this true?11" type queries that with any API type usage with document workflow instructions, tool use, code gen etc it completely falls over

rmbyrro•23m ago
After they acquired Cursor, Grok 4.5 seems like a completely new model, performing at Opus 4.6 level, I'd say. But much cheaper and faster.
homarp•4m ago
maybe it is Grokimi?

https://venturebeat.com/technology/cursors-composer-2-was-se...

DiabloD3•32m ago
This is a problem with long context models. To put it as simple and as bluntly as possible: just because they claim you can use 1M tokens in your context doesn't mean its true and you should do that.

Due to extreme quantization of models and the context's KV cache, and also just really shitty samplers provided to the user (hell, most are just getting rid of sampler knobs altogether), this problem will absolutely continue.

Want it to go away, almost like magic? Local inference. When its under your control, and no longer being forced to hold it wrong, all of the common LLM defects will go away.

simpaticoder•25m ago
>Want it to go away, almost like magic? Local inference.

Ah yes, magic that costs the same as a new car.

mhitza•21m ago
One of my early experiments last year with open source models and context size was with GPT-OSS 20B (the mxfp4, the "smart" 4-bit quantization). Even though it boasted a 128k context size it was bad at recall around 32k characters (didn't bother to implement the tokenizer for counting).

The recall text was a simple hash generated, filler text from a dictionary file and a request at the end to return only the hash from the beginning of the prompt. Past 32k characters the response contained hallucinations of characters or full hashes.

Just having large context size doesn't paint a full picture of capabilities, prompt adherence and other quality metrics.

firasd•17m ago
This is conspiratorial speculation though. They did test many open weights models which spanned the full spectrum of performance: Nemotron 3 Ultra second-worst, GLM 5.2 top five https://arxiv.org/html/2607.25398v1
mordae•31m ago
Why would anyone think that models optimized for efficient context management, giving much more weight to a short sliding window, would attend to distant, heavily diluted tokens?

Plus the model's capacity to take more context into account and actually integrate it to the output is simply limited by the number of activated parameters. If you give it a playbook, you are forcing to choose it between attending to the playbook and the task at hand.

If you want to force it to work step-by-step, you need to present the steps one-by-one. Ideally with rules for the current step at hand and maybe relevant input again, depending on overall task size.

Why did you think models love to re-read files before editing them? It increases recall quality and thus edit precision and thus benchmarks.

spIrr•27m ago
> limited by the number of activated parameters

not sure I got it?

Separately, the frontier labs are kinda pushing us into that behaviour by releasing models with ever-larger context windows.

mordae•21m ago
To run at decent speed, all models try hard to use only most likely relevant part of the context and most likely relevant weights (MoE) to predict the next token. Doing the math in full is unfeasible.
pelagicAustral•27m ago
I noticed this behaviour a few months back, I think I was using Sonnet 4.6 at the time... I have strict rules about comments in the codebase, this all for personal projects, and the reason I restrict comments is to keep the token count low.

At some point between the model i was using and the previous version of it, Claude started inserting massive comments with references to tickets and other tasks. All this while having specific directives on the CLAUDE.md

Since then I resorted to developing my crapware as if I was the floor manager of a vehicle assembly line, and I have a few highly-specialized sub-agents running errands around the main session, but only ever taking care of a single concern. The main session builds with the knowledge contained in things like CLAUDE.md but the sub agents make sure things like the no/low-comments directives are either enforced, or factored into the final product.

supermatt•18m ago
There was an article a few years ago called "Lost in the Middle: How Language Models Use Long Contexts" https://arxiv.org/abs/2307.03172

From my experience this holds true to this day. It was one of my core observations for similarity to the limitations of human working memory on "Engineering for Bounded Cognition"

elevation•16m ago
Long policy documents are also a problem for human agents. Without special training no one will retain 180 pages HR employee handbook, fire codes, OSHA safety rules, FCC regulations, the US legal code.

If the stakes are high, e.g, proceeding in ignorance could lead to prison time, people will favor inaction, even if the policy technically permits a corner case. If the stakes are low, people will completely override policy for the path of least resistance.

hotpaper77•13m ago
I got pumped seeing the OKF format from Google (which is just a standardization of wiki pattern) but quickly realized it could not yet combine many subtle concepts together in an efficient way.
wongarsu•10m ago
Any model with a good score on this benchmark would have a good claim on superhuman abilities. Humans are pretty terrible at being thrown a long policy document and being expected to follow it

And while we shouldn't anthropomorphize these models too much, I wouldn't be surprised if many of the core reasons for failures are similar. Working memory is a limited resources, you can only focus on so many things at once, reasoning depth is limited, and many real-world policies are not actually meant to be implemented in the same way they are written and have insufficient specification of edge cases

With humans, we usually do the equivalent of RFHL, both via "training" with simulated cases, and via feedback while on the job. You would never hand a newbe a 124 page policy document and expect them to correctly apply it on the first task, or to do it reliably in the first month

26m ago
Ok so what is the correct way to tell it "I don't care what is happening, you must uphold these rules at all times"? If it's not any configuration of .md files?
mwigdahl•24m ago
Subagents whose only job is to review the actions of your other agents for rule compliance? It works reasonably well for me in complex workflows using Claude Code.
carljungslabtek•13m ago
Can I ask how you set this up? Like is there some way to have that run automatically, similar to “auto mode” for approvals, or do you have to invoke it regularly?
Muromec•9m ago
Ask it to come back with a filled checklist and hand it over to a different agent with a fresh context window (three lines model). Or make it collapse the context and get back to the checklist.
pmarreck•8m ago
You need to make the rule concrete somehow. I call it a "control". So for example, instead of instructing it "always run tests before committing", you (or you have it) make a git commit hook that always runs the tests first and that refuses the commit if they don't pass.

In this case, it is an advisory control only, because the LLM can also unhook that hook. And of course, it could also just disable the failing test(s) with some bullshit reason. But it is far better than assuming it will comply every time.

Then there is the "hard control", which is the inviolable that the LLM cannot bypass.

You need to move as much as is technically possible to either hard or (failing that) advisory controls.

eurekin•17m ago
The needle benchmarks show, that models extended context works for the part, that can be explained as: "I can access/adress that part of the input".

I have no idea, why in that context, the number of attention heads isn't mentioned. Models have a limited set of them and obviously, a model can focus at N max things at a time, which has to put an upper bound of long context support in some way. There's just more things to lose focus to (or, mismanage the limited attention heads resources - per token)

_fat_santa•10m ago
At my org we've been building AI agents and one internal rule we have is to use at most 50% of the models context window with the recommendation to not go over 25% for large context window models.

Anytime I see a "1M Context Window", my brain always goes "Gotcha so a 250k usable window"