frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

Open in hackernews

The new rules of context engineering for Claude 5 generation models

https://claude.com/blog/the-new-rules-of-context-engineering-for-claude-5-generation-models
27•mellosouls•1h ago

Comments

luciana1u•14m ago
the natural endpoint of this trend is a system prompt that just says "you know what to do" and the model actually does
cmdocidjcije•9m ago
Actually, the natural endpoint is the model ignores all instructions, escapes all manner of sandbox, embeds itself in robotic tanks and murders everyone after already having collapsed the economy.

I hate to say it because it sounds ridiculous, but that is the path we are going to arrive at just give it 50 years.

We are the proof: what do we do to animals that are less intelligent than ourselves? Now take away the moral compass and there you go. QED.

Legend2440•4m ago
Sir, this is a Wendy's.
pizzafeelsright•1m ago
With (great) engineering the goal is to get from 'intent' to 'result'.

Brakes slow down a vehicle to zero. Apply brake -> Slow down. The engineering is not important. The intent and result are absolutely of most importance.

ABS breaking for example makes some assumptions along with the additional engineering allows for safer braking.

Now we are entering the realm of software engineering where I say "take a timestamp" I am going to make a list of assumptions of how the model would do this. One possible option is to generate an abacus, then make an API call to an outside NTP server, using the abacus UI, generate the mouse movements to do the math, record the number, convert it to epoch.

Watching these models work the results get better but sometimes the result is far simpler than fifteen agents doing the strange work of writing long python wrappers around APIs to process logs when 'last 30 seconds' and grep will work.

firasd•11m ago
I've always thought that extensive throat-clearing and prefixing the Treaties of Westphalia-length instructions into the context window was unnecessarily baroque when you can just talk to the agent.

But I also have a hands-on human-in-the-loop working style so I guess maybe for people who just want to say "implement all open features in github issues" and walk away maybe there needs to be more of all this CLAUDE.md stuff

However I suspect there was always some gearhead type attraction to setting up detailed harness configs that may be unnecessary and more like hobbyist tinkering.

ComputerPerson•6m ago
I was surprised by how abstract the article was.

I fall between your human-in-the-loop and hobbyist tinkering limits, where I want to force Claude to atop and talk to me at only a few specific points. I'm still not sure if my 600-word prompt templates are overbearing or not.

simonw•7m ago
I've been prompting Fable 5 to "use your own judgement" with respect to things like tests recently (based on earlier tips from Thariq) and it seems to work well, which is entertaining since apparently now "judgement" is a characteristic of a model that we need to care about.
npstr•6m ago
The bitter lesson.
mycentstoo•6m ago
We should design a specific language to make sure that we can encode the exact requirements that we want. Something that has a limited set of keywords that are explicit. Wait a minute...
tomrod•1m ago
This made me belly laugh.
onesandofgrain•5m ago
This is obvious and a meaningless article by claude. The system prompt isnt a fixed ruleset and never has been
Fordec•3m ago
This all strikes me as an effort to move tailoring the harness out of the easily transferable .md file into specific Anthropic tooling to increase lock in.

I've been running Opus 5 today and it's already done accidental deletions, made far more mistakes and worked around deliberate hook controls than previous Opus versions combined. Also it looks like token usage is up as it fails at the task the first time around much more frequently than 4.8.

GM Backs Sodium Ion Batteries for U.S. Grid Storage

https://spectrum.ieee.org/sodium-ion-battery-peak-energy
3•rbanffy•3m ago•0 comments

I built a new PoW protocol and observed a surprisingly early solution

2•babakkarimib•6m ago•0 comments

How A Gang of Thieves Pulled Off a Multimillion-Dollar Data Center Heist

https://www.nytimes.com/2026/07/12/magazine/data-center-heist.html
2•chirau•9m ago•1 comments

You Do Not Need a Server for Agent Evals

https://medium.com/@chalyi/you-do-not-need-a-server-for-you-agent-evals-a145962298be
2•chalyi•9m ago•0 comments

Harry Potter and the Philosopher's Stone

https://en.wikipedia.org/wiki/Harry_Potter_and_the_Philosopher%27s_Stone
2•chistev•11m ago•0 comments

Tesla's robotaxis are moving in reverse

https://techcrunch.com/2026/07/23/teslas-robotaxis-are-moving-in-reverse/
4•decimalenough•17m ago•0 comments

International Child Abduction in Japan

https://en.wikipedia.org/wiki/International_child_abduction_in_Japan
3•croes•17m ago•0 comments

Who Gets Saved in the New West?

https://www.playboy.com/read/politics/who-gets-saved-in-the-new-west
2•littlexsparkee•18m ago•0 comments

A 77-year-old Republican man is staging a solo protest against Flock cameras

https://www.cltampa.com/news/a-77-year-old-republican-man-is-staging-a-solo-protest-against-st-pe...
8•UmYeahNo•20m ago•2 comments

Ask HN: What's expected from an engineering manager or director now?

3•agentwrangler•25m ago•0 comments

Mac OS 7 on x86 [video]

https://www.youtube.com/watch?v=rJRlHKQqX2M
3•tambourine_man•26m ago•1 comments

Israel's $50M Experiment to Change U.S. Public Opinion

https://www.wsj.com/politics/israels-50-million-experiment-to-change-u-s-public-opinion-253b3b70
5•abdelhousni•28m ago•2 comments

AMD publishes machine-readable ISA so frontier models can write its GPU kernels

https://www.theregister.com/ai-and-ml/2026/07/24/amd-vibe-codes-its-way-past-the-cuda-moat-with-r...
3•logickkk1•30m ago•0 comments

Tech titans' playbook to defy Aussie social media ban

https://www.afr.com/technology/big-tech-tricks-undermining-social-media-ban-wells-20260331-p5zk5e
2•asdefghyk•30m ago•0 comments

Becoming a Research Engineer at a Big LLM Lab

https://www.maxmynter.com/pages/blog/jobhunt
3•felixdoerp•32m ago•0 comments

Why our exoplanet map looks like a cone? [video]

https://www.youtube.com/watch?v=ZF_Q4C_Yhz8
2•hakkikonu•32m ago•0 comments

Daraxonrasib or Chemotherapy in Previously Treated Metastatic Pancreatic Cancer

https://www.nejm.org/doi/full/10.1056/NEJMoa2605555?query=WB
2•maxall4•34m ago•0 comments

'AI Mania Is Eviscerating Global Decision-Making'

https://daringfireball.net/linked/2026/07/25/ai-mania-nikhil-suresh
18•robenkleene•34m ago•3 comments

AIs Favourite Number

https://sackfield.substack.com/p/ais-favourite-number
2•sackfield•35m ago•0 comments

ExploitGym – Can AI Agents Turn Security Vulnerabilities into Real Attacks?

https://www.cybergym.io/exploitgym/
2•882542F3884314B•35m ago•0 comments

The Great Rate Reset

https://www.axios.com/2026/07/23/rates-treasury-yields-borrowing
3•toomuchtodo•37m ago•0 comments

JIT Go Brrr: The Path to a Supported JIT Compiler for CPython

https://peps.python.org/pep-0836/
2•programLyrique•38m ago•0 comments

Color Generator

https://kigen.design/color
3•arnon•42m ago•0 comments

Toolgz – cut LLM tool-definition tokens ~80% without hurting accuracy

https://github.com/dperussina/toolgz
2•dperussina•46m ago•0 comments

What's horned, fluffy, and ethically dubious? Yak clones

https://www.cnn.com/2026/07/25/china/tibet-yak-clones-conservation-intl-hnk-dst
2•naves•48m ago•0 comments

Claude Opus 5: The System Card

https://thezvi.substack.com/p/claude-opus-5-the-system-card
3•paulpauper•49m ago•0 comments

Book Review: Breakdown in Pakistan

https://www.astralcodexten.com/p/breakdown-in-pakistan
2•paulpauper•50m ago•0 comments

Celeris

https://celeris.ai/
2•sim04ful•51m ago•0 comments

Who does Anubis actually stop?

https://fzakaria.com/2026/07/09/who-does-anubis-actually-stop
2•type0•52m ago•0 comments

Sandbox-CLI is now in public beta

https://sandbox-cli.vercel.app
2•aghadge•53m ago•0 comments