frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

fp.

Show HN: A persistent fact store for AI world simulations

https://github.com/clay-good/worldify
1•hireclay•38s ago•0 comments

Implementing MQTT 5 in Go (Part II): QoS and Session Management

https://medium.com/@MonsieurTib/implementing-mqtt-5-in-go-a-deep-dive-into-client-design-part-ii-...
1•MrTib•5m ago•0 comments

Indie bookstores are making a triumphant comeback

https://www.fastcompany.com/91461983/indie-bookstores-are-making-a-shocking-triumphant-comeback
2•atlasunshrugged•7m ago•0 comments

Lua 5.5.0 Now Available

https://groups.google.com/g/lua-l/c/jW6vCnhVy_s
2•ligurio•10m ago•0 comments

Show HN: withcodemode – Proxy any Remote MCP server to run under codemode

https://github.com/jaw9c/withcodemode
2•joshwarwick15•11m ago•0 comments

ByteDance's valuation hits US$500B as US survival plan advances

https://www.scmp.com/tech/big-tech/article/3337343/tiktok-owner-bytedances-valuation-hits-us500-b...
2•mfiguiere•11m ago•0 comments

Vince Zampella, co-creator of Call of Duty video game series, dies aged 55

https://www.theguardian.com/us-news/2025/dec/22/vince-zampella-co-creator-of-call-of-duty-video-g...
2•vinnyglennon•12m ago•1 comments

2025 in Review – Vicki Boykis

https://vickiboykis.com/2025/12/22/2025-in-review/
2•Anon84•13m ago•0 comments

The Hum

https://en.wikipedia.org/wiki/The_Hum
2•fi-le•13m ago•0 comments

Google Buys Data Center Company for $4.75B

https://www.nytimes.com/2025/12/22/technology/google-intersect-data-centers.html
2•xnx•14m ago•0 comments

From a For-Loop to Transformers

https://python2llms.org/
1•yegortk•15m ago•0 comments

'I'm leaving [Ireland]': Migrant workers call for action in wake of attacks

https://www.thejournal.ie/indian-attacks-ireland-6789412-Aug2025/
3•TMWNN•16m ago•1 comments

Psychometric Jailbreaks Reveal Internal Conflict in Frontier Models

https://arxiv.org/abs/2512.04124
2•berlam•17m ago•0 comments

Oracle's AI Bet Is Facing Its First Stress Test

https://www.profgmarkets.com/p/oracle-s-ai-bet-is-facing-its-first-stress-test
3•andyjohnson0•18m ago•0 comments

I want to fix Durable Execution

https://github.com/maxnorth/rhythm
3•mnorth•18m ago•2 comments

Design is more than code

https://linear.app/now/design-is-more-than-code
2•samsolomon•19m ago•0 comments

Boys shared AI-generated, nude images of her. After a fight, she was expelled

https://apnews.com/article/school-deepfake-nude-ai-cyberbullying-0ead324241cf390e1a7f3378853f23cb
3•healsdata•21m ago•0 comments

QuickCurrency – Free currency converter with no sign-up required

https://www.quickcurrency.net
2•DDARJEAN•25m ago•0 comments

Show HN: G=(hbar*c*2*(1+alpha/3)^2)/(m_p^2*4^64) ≈ 6.6742439706e-11 (8 ppm)

2•albert_roca•25m ago•1 comments

BlazorOcticons – The easiest way to use GitHub Octicons in your Blazor apps

https://blazorocticons.net/
2•elv1s42•29m ago•1 comments

Framework Laptop 16 DIY OCuLink x8 expansion module adds desktop GPU support

https://liliputing.com/framework-laptop-16-diy-oculink-x8-expansion-module-brings-desktop-gpu-sup...
4•PaulHoule•30m ago•0 comments

Coding Rust with Claude Code and Codex

https://tigran.tech/coding-rust-with-claude-code-and-codex/
2•tigranbs•31m ago•0 comments

The Year in Books: 2022

https://notes.eatonphil.com/2023-01-12-year-in-books.html
3•ibobev•32m ago•0 comments

Kimchi, made in China: how South Korea's national dish is being priced out

https://www.theguardian.com/world/2025/dec/22/kimchi-south-korea-national-dish-priced-out-china-e...
2•n1b0m•32m ago•0 comments

Capital One Is Losing Its Travel Dominance: A Review of the Discover Debit Cards

https://testinginprod.blog/capital-one-is-losing-its-travel-dominance-a-review-abroad-of-their-ne...
4•charliebwrites•32m ago•0 comments

In Pursuit of Clancy Sigal (2021)

https://yalereview.org/article/in-pursuit-of-clancy-sigal
3•dang•34m ago•0 comments

Trying to Fit Exponential Data

https://www.johndcook.com/blog/2025/12/22/fit-exponential-data/
2•ibobev•34m ago•0 comments

Please give some tips to shorten the Linux device name of NVMe (2023)

https://old.reddit.com/r/linuxquestions/comments/1biouhv/please_give_some_tips_to_shorten_the_nam...
2•transpute•36m ago•0 comments

Infinite Ethics [pdf]

https://nickbostrom.com/ethics/infinite.pdf
3•ChadNauseam•37m ago•0 comments

Dad's Fitness May Be Packaged and Passed Down in Sperm RNA

https://www.quantamagazine.org/how-dads-fitness-may-be-packaged-and-passed-down-in-sperm-rna-2025...
4•ibobev•37m ago•0 comments
Open in hackernews

GLM-4.7: Advancing the Coding Capability

https://z.ai/blog/glm-4.7
84•pretext•2h ago

Comments

esafak•50m ago
The terminal bench scores look weak but nice otherwise. I hope once the benchmarks are saturated, companies can focus on shrinking the models. Until then, let the games continue.
CuriouslyC•42m ago
We're not gonna see significant model shrinkage until the money tap dries up. Between now and then, we'll see new benchmarks/evals that push the holes in model capabilities in cycles as they saturate each new round.
lanthissa•28m ago
isn't gemini 3 flash already model shrinkage that does well in coding?
hedgehog•25m ago
Smaller open-weights models are also improving noticeably (like Qwen3 Coder 30B), the improvements are happening at all sizes.
cmrdporcupine•20m ago
Devstral Small 24b looks promising as something I want to try fine tuning on DSLs, etc. and then embedding in tooling.
bigyabai•27m ago
It's a good model, for what it is. Z.ai's big business prop is that you can get Claude Code with their GLM models at much lower prices than what Anthropic charges. This model is going to be great for that agentic coding application.
theshrike79•10m ago
z.ai models are crazy cheap. The one year lite plan is like 30€ (on sale though).

Complete no-brainer to get it as a backup with Crush. I've been using it for read-only analysis and implementing already planned tasks with pretty good results. It has a slight habit of expanding scope without being asked. Sometimes it's a good thing, sometimes it does useless work or messes things up a bit.

cmrdporcupine•27m ago
Running it in Crush right now and so far fairly impressed. It seems roughly in the same zone as Sonnet, but not as good as Opus or GPT 5.2.
XCSme•25m ago
Funny how they didn't include Gemini 3.0 Pro in the bar chart comparison, considering that it seems to do the best in the table view.
jychang•11m ago
Also, funny how they included GPT-5.0 and 5.1 but not 5.2... I'm pretty sure they ran the benchmarks for 5.0, then 5.1 came out, so they ran the benchmarks for 5.1... and then 5.2 came out and they threw their hands up in the air and said "fuck it".
XCSme•4m ago
I didn't even notice that, I assumed it was the latest GPT version.
Tiberium•12m ago
The frontend examples, especially the first one, look uncannily similar to what Gemini 3 Pro usually produces. Make of that what you will :)

EDIT: Also checked the chats they shared, and the thinking process is very similar to the raw (not the summarized) Gemini 3 CoT. All the bold sections, numbered lists. It's a very unique CoT style that only Gemini 3 had before today :)

reissbaker•11m ago
I don't mind if they're distilling frontier models to make them cheaper, and open-sourcing the weights!
jtrn•10m ago
My quickie: MoE model heavily optimized for coding agents, complex reasoning, and tool use. 358B/32B active. vLLM/SGLang only supported on the main branch of these engines, not the stable releases. Supports tool calling in OpenAI-style format. Multilingual English/Chinese primary. Context window: 200k. Claims Claude 3.5 Sonnet/GPT-5 level performance. 716GB in FP16, probably ca 220GB for Q4_K_M.

My most important takeaway is that, in theory, I could get a "relatively" cheap Mac Studio and run this locally, and get usable coding assistance without being dependent on any of the large LLM providers. Maybe utilizing Kimik2 in addition. I like that open-weight models are nipping at the feet of the proprietary models.

embedding-shape•8m ago
> Supports tool calling in OpenAI-style format

So Harmony? Or something older? Since Z.ai also claim the thinking mode does tool calling and reasoning interwoven, would make sense it was straight up OpenAI's Harmony.

> in theory, I could get a "relatively" cheap Mac Studio and run this locally

In practice, it'll be incredible slow and you'll quickly regret spending that much money on it instead of just using paid APIs until proper hardware gets cheaper / models get smaller.

reissbaker•4m ago
s/Sonnet 3.5/Sonnet 4.5

The model output also IMO look significantly more beautiful than GLM-4.6; no doubt in part helped by ample distillation data from the closed-source models. Still, not complaining, I'd much prefer a cheap and open-source model vs. a more-expensive closed-source one.

gigatexal•6m ago
Even if this is one or two iterations behind the big models Claude or openai or Gemini it’s showing large gains. Here’s hoping this gets even better and better and I can run this locally and also that it doesn’t melt my PC.
larodi•3m ago
From my limited exposure to these models, they seem very very very promising.