frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

Open in hackernews

Fable 5 vs. GPT-5.6 Sol on an NP-Hard Problem: Does /goal help?

https://charlesazam.com/blog/fable-5-gpt-5-6-sol-goal/
54•couAUIA•2h ago

Comments

couAUIA•2h ago
A deepdive on the /goal effect on a problem literally made for this.
tyleo•1h ago
The chart at the top is somewhat confusing. It says, “lower is better” but the y-axis is inverted! So visually higher in the chart is better but lower in terms of # value.
couAUIA•1h ago
well thank you so much for this
o10449366•1h ago
/goal has replaced plan mode for me. This is the pattern I use for 95% of my AI work now:

1. Read X feature of Y and tell me when you fully understand it (if there's any detail missing in the summary, repeat until the context is primed)

2. What time is it?

3. /goal Spend X minutes from $time writing a technical design doc on $feature. There must not be any vague language or ambiguity in the document. Read carry_forward_requirements.md and testing_best_practices.md and explicitly incorporate them into the document you write. The document should be executable for a contextless implementer when done and include specific code and document references and changes needed. Spend the full X minutes working on and reviewing this document - do not quit early and wait

Even just spending 10 minutes forcing GPT to write a design doc results in much more robust plans than plan mode, in my experience, and saves time I would spend iterating on the initial plan mode draft anyway.

embedding-shape•1h ago
> 3. /goal Spend X minutes from $time writing a technical design doc on $feature.

Hmm, I feel like this is akin to making a recursive function have a exit condition not based on what it actually did/found, but based on how long time it took.

I'm always using /goal with explicit goals that the agent needs to achieve. Time-bounding them wouldn't make sense, I want something specific done regardless of how long time it takes.

So instead I'd put goals on what the design/architecture needs to achieve, and for the model to continuously check the outcome against these, then finish when everything is achieved. Doesn't really matter if it takes 10 minutes or 10 hours, which for me is a bit the point of /goal in the first place, otherwise I'd just use the agent normally.

illliillll•1h ago
Well, I’ve been having 5.6 sol work on tasks like “find every OTA app on the internet”

I find explicit time bounds are useful for tasks like this, otherwise the LLM will almost certainly return too early.

lonelyasacloud•10m ago
Leaving to run unhindered makes sense when the agents understand both the complete domain context and can be supplied with a strong exit criteria that they understand well.

Otherwise time boxing is both going to help stop entities from wandering off into the weeds. And also communicate expectations from the commissioner about the expected effort levels and output quality requirements.

Empirically in human world, get very different results when an employee, particularly a junior, is asked to spend 1/2 a day on a work package, a week or are left completely to their own devices.

bob1029•39m ago
The hypothesis generation phase is looking like the most critical part of having an agent reliably hit your targets.

Simply starting in the correct part of the search space is probably the biggest predictor of success. Forcing one big loop to fight its way through all the hypotheticals from zero looks like a dead end for many practical scenarios, regardless of how powerful the model is. I think you could draw some analogies to humans here.

I have found that delegating deep research to a simple tool call is the best way to ground the agent in complex domains. If you make the main agent loop carry the weight of this research, it's going to do a really shitty job because of how the RLHF tries to preserve context and get an answer to the user quickly. As a tool, you may find the agent invokes multiple rounds of research consecutively without realizing it has incurred billions of tokens of consumption. Many of the tokens are wasted when generating independent hypotheses and subsequently investigating them, but the point is that you sampled 10-100x search space before getting serious about mutating the environment. The tradeoff seems worth it in a lot of cases. Correctness >> Time >> Money.

techpression•1h ago
I love that we have this on one hand and me cleaning up catastrophic CSS made by Sol on the other. Then again, maybe CSS is the ultimate benchmark.
baq•1h ago
CSS is the reason I refuse to do any frontend work except FE infra and I know I’m not alone here, soooo yes I guess?
cwmoore•46m ago
I do not know the whole picture, but if you are asking for blind one-shot CSS, you might benefit from wiring the model to take screenshots of various end-browsers and discuss them as you iterate.

Offering freelance estimates for CSS design changes before frameworks were around was a problem.

andai•1h ago
Results seem mostly noise to me. One eval per model, in a large problem space (i.e. a problem which requires many attempts to solve well).
couAUIA•1h ago
Yes I agree, but I actually did a lot more runs, with different prompts, different times ect... And each time /goal had a small or insignificant impact
tantalor•1h ago
What is /goal?
swader999•1h ago
On Claude if you start with that, it won't stop until it achieves or exhausts your prompt. It feels like "here's your mission, go do it". I use it a few times a week.
ryan_n•1h ago
The llm runs in a loop until it meets a condition (the goal).
Tenoke•44m ago
Claude seems to forget what you tell it in very long work sessions (things that take weeks to develop), no matter how many times you tell it which part is extra important. I dont use goal (I guess I should), but presumably it makes it actually remember the most important instruction. I believe this here is about shorter sessions where the issue doesn't crop up as much.
jswny•37m ago
Unfortunately I’ve used Claude and GPT models for a long time in a variety of harnesses and I agree with you and I think it’s the compaction.

For some reason, codex compaction is like black magic. I’ve never felt like I can just one one continuous thread with other models, Claude I carefully curate when I compact

semiquaver•30m ago
Unlike other model/harness pairs, codex+gpt also passes an opaque encrypted artifact speculated to be an embedding representing the conversation back to the successor generation which is “denser” or at least higher fidelity than summarized text.
brookst•3m ago
Compacting at all is a mistake. With 1m context window there is no reason for a single task to require compaction.

Much better to spend tokens breaking the task into chunks, documenting and storing them durably, then executing each one in clean context and just /clear after.

It’s a similar concept to compaction, just planned in advance. Much much more effective, and doesn’t burn tokens and time (“wall-clock”, Claude) doing the compaction.

IceDane•22m ago
Stop using compaction. Stop treating sessions as something you keep alive for weeks. Set up a proper workflow for handing off work and start handing off well before maximum context is hit. It's less efficient, slower and more costly to constantly be sitting at high context and compaction is just not good.
esperent•16m ago
This is one of the advantages of pi. I made a /protect command that protects the message from compaction. I also protect skills automatically.

So for long running tasks I'll do

    /protect your goal is...

GPT-5.6 used a prompt to close a 30-year gap in convex optimization

https://old.reddit.com/r/math/comments/1uxj3cy/after_openais_cdc_proof_announcement_gpt56_used_a/
69•mbustamanter•54m ago•22 comments

LG monitors silently install software through Windows Update without consent

https://videocardz.com/newz/lg-monitors-silently-install-software-through-windows-update-without-...
391•baranul•3h ago•193 comments

Fable 5 vs. GPT-5.6 Sol on an NP-Hard Problem: Does /goal help?

https://charlesazam.com/blog/fable-5-gpt-5-6-sol-goal/
54•couAUIA•2h ago•22 comments

Regressive JPEGs

https://maurycyz.com/projects/bad_jpeg/
475•vitaut•10h ago•46 comments

The Computer at the Bottom of a Canal

https://negroniventurestudios.com/2026/07/18/the-computer-at-the-bottom-of-a-canal/
59•Kudos•5h ago•10 comments

AWS: Inaccurate Estimated Billing Data – $1.7 billion

1229•nprateem•1d ago•718 comments

Is this the end of the once-mighty GoPro?

https://amateurphotographer.com/latest/photo-news/going-going-gone-is-this-the-end-of-the-once-mi...
13•aanet•3d ago•15 comments

Reviving a 15-year-old netbook with Arch Linux

https://parksb.github.io/en/article/41.html
156•parksb•3d ago•94 comments

Thanks HN for 15 years of support and helping me find my life's work

686•nicholasjbs•20h ago•80 comments

Qubes OS Security in the Public Record

https://arxiv.org/abs/2607.14587
36•sciences44•5h ago•5 comments

First atmosphere found on Earth-like planet in habitable zone of distant star

https://www.bbc.com/news/articles/cy4kdd1e0ejo
479•neversaydie•23h ago•282 comments

Show HN: IKEA Complexity Index

https://ikea.greg.technology/
101•gregsadetsky•10h ago•53 comments

Waldi: A quiet place to write, and to be read

https://github.com/waaldev/waldi
31•waaldev•6d ago•15 comments

The Zilog Z80 has turned 50

https://goliath32.com/blog/z80.html
244•st_goliath•18h ago•97 comments

TP-Link Kasa cameras leaked home GPS via unauthenticated UDP for 6 years

https://github.com/BadChemical/IoT-Vulnerability-Research-Public/blob/main/TP-Link_Kasa_EC71/Kasa...
162•BadChemical•16h ago•60 comments

In-toto: A framework to secure the integrity of software supply chains

https://in-toto.io/
49•Erenay09•1d ago•7 comments

Steam Machine: Between 12k and 15k Units Sold per week

https://boilingsteam.com/steam-machine-between-10k-and-15k-sold-per-week/
46•ekianjo•2h ago•26 comments

Learning a few things about running SQLite

https://jvns.ca/blog/2026/07/17/learning-about-running-sqlite/
281•surprisetalk•20h ago•75 comments

What AI did to stackoverflow in a graph

https://data.stackexchange.com/stackoverflow/query/1953768#graph
114•secretslol•2h ago•127 comments

Kimi K3, and what we can still learn from the pelican benchmark

https://simonwillison.net/2026/Jul/16/kimi-k3/
365•droidjj•23h ago•192 comments

I started a “dirt notebook”

https://pinewind.bearblog.dev/i-started-a-dirt-notebook/
86•herbertl•12h ago•75 comments

Stenchill: 3D Printable Solder Paste Stencil Generator

https://www.stenchill.com/en/
67•radeeyate•13h ago•19 comments

Why do AI company logos look like buttholes? (2025)

https://velvetshark.com/ai-company-logos-that-look-like-buttholes
303•miniBill•2h ago•94 comments

Static search trees: 40x faster than binary search (2024)

https://curiouscoding.nl/posts/static-search-tree/
150•lalitmaganti•17h ago•11 comments

Revisiting Yliluoma's ordered dither algorithm

https://30fps.net/pages/revisiting-yliluoma-2/
17•ibobev•4d ago•1 comments

Open Book Touch: open-source e-reader

https://www.crowdsupply.com/oddly-specific-objects/open-book-touch
143•surprisetalk•17h ago•45 comments

DrDroid (YC W23) Is Hiring

https://www.ycombinator.com/companies/drdroid/jobs/w45QcNV-product-engineer-assignment-mandatory
1•TheBengaluruGuy•12h ago

An Update on Igalia's Layer Based SVG Engine in WebKit (Reducing Layer Overhead)

https://blogs.igalia.com/nzimmermann/posts/2026-07-14-lbse-conditional-layers/
62•bkardell•4d ago•4 comments

Battery packs: Let's talk about crates, baby

https://smallcultfollowing.com/babysteps/blog/2026/07/15/battery-packs/
37•MeetingsBrowser•1d ago•23 comments

Vāgdhenu: A Sanskrit Chanting TTS System

https://prathosh.in/vagdhenu/
196•subinalex•4d ago•48 comments