frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

Graphene OS: a security-enhanced Android build

https://lwn.net/SubscriberLink/1030004/898017c7953c0946/
186•madars•6h ago•82 comments

Scientists may have found a way to eliminate chromosome linked to Down syndrome

https://academic.oup.com/pnasnexus/article/4/2/pgaf022/8016019
224•MattSayar•6h ago•170 comments

Inter-Planetary Network Special Interest Group

https://www.ipnsig.org
115•OhMeadhbh•7h ago•29 comments

Positron – A next-generation data science IDE

https://positron.posit.co/
126•amai•3d ago•49 comments

I wasted weeks hand optimizing assembly because I benchmarked on random data

https://www.vidarholen.net/contents/blog/?p=1160
250•thunderbong•3d ago•76 comments

AMD CEO sees chips from TSMC's US plant costing 5%-20% more

https://www.bloomberg.com/news/articles/2025-07-23/amd-ceo-su-sees-chips-from-us-tsmc-plant-costing-5-to-20-more
282•mfiguiere•1d ago•461 comments

There is no memory safety without thread safety

https://www.ralfj.de/blog/2025/07/24/memory-safety.html
285•tavianator•12h ago•265 comments

Alto turns your Apple Notes into a website

https://alto.so/
39•colinprince•3h ago•14 comments

A GPU Calculator That Helps Calculate What GPU to Use

https://calculator.inference.ai/
52•chlobunnee•5h ago•15 comments

Air Force unit suspends use of Sig Sauer pistol after shooting death of airman

https://www.nhpr.org/nh-news/2025-07-23/sig-sauer-pistol-air-force-shooting-death
130•duxup•9h ago•225 comments

PSA: SQLite WAL checksums fail silently and may lose data

https://avi.im/blag/2025/sqlite-wal-checksum/
246•avinassh•13h ago•111 comments

RE#: High performance derivative-based regular expression matching (2024)

https://arxiv.org/abs/2407.20479
32•fanf2•3d ago•5 comments

Visa and Mastercard: The global payment duopoly (2024)

https://quartr.com/insights/edge/visa-and-mastercard-the-global-payment-duopoly
258•bilekas•6h ago•148 comments

New Aarch64 Back End

https://ziglang.org/devlog/2025/#2025-07-23
75•Bogdanp•6h ago•24 comments

Use Your Type System

https://www.dzombak.com/blog/2025/07/use-your-type-system/
235•ingve•13h ago•242 comments

Revisiting Moneyball

https://djpardis.medium.com/revisiting-moneyball-074fc2435b07
68•sebg•6h ago•26 comments

Open Source Maintenance Fee

https://github.com/wixtoolset/issues/issues/8974
235•AndrewDucker•15h ago•167 comments

Information Warfare

https://yuxi.ml/sketches/posts/info-warfare/
14•mesaoptimizer•3d ago•3 comments

How Anthropic teams use Claude Code

https://www.anthropic.com/news/how-anthropic-teams-use-claude-code
120•yurivish•2h ago•93 comments

Covers as a way of learning music and code

https://ntietz.com/blog/covers-as-a-way-of-learning/
137•zdw•3d ago•75 comments

Vet is a safety net for the curl | bash pattern

https://github.com/vet-run/vet
182•mooreds•15h ago•168 comments

Intel CEO Letter to Employees

https://morethanmoore.substack.com/p/intel-ceo-letter-to-employees
209•fancy_pantser•7h ago•372 comments

Why concatenative programming matters (2012)

http://evincarofautumn.blogspot.com/2012/02/why-concatenative-programming-matters.html
52•azhenley•3d ago•13 comments

Low-Temp 2D Semiconductors: A Chipmaking Shift

https://spectrum.ieee.org/cdimensions-2d-semiconductors
6•rbanffy•3d ago•2 comments

Bus Bunching

https://www.futilitycloset.com/2025/07/12/bus-bunching/
54•surprisetalk•4d ago•59 comments

Writing is thinking

https://www.nature.com/articles/s44222-025-00323-4
283•__rito__•3d ago•118 comments

Mwm – The smallest usable X11 window manager

https://github.com/lslvr/mwm
141•daureg•3d ago•58 comments

Show HN: Nia – MCP server that gives more docs and repos to coding agents

https://www.trynia.ai/
55•jellyotsiro•13h ago•44 comments

UK: Phone networks down: EE, BT, Three, Vodafone, O2 not working in mass outage

https://www.the-independent.com/tech/ee-bt-three-vodafone-o2-down-phone-networks-outage-latest-b2795260.html
198•oger•14h ago•87 comments

The POSIX specification of vi

https://pubs.opengroup.org/onlinepubs/9799919799/utilities/vi.html
65•exvi•3d ago•24 comments
Open in hackernews

Kimi-K2 Tech Report [pdf]

https://github.com/MoonshotAI/Kimi-K2/blob/main/tech_report.pdf
82•swyx•3d ago

Comments

dang•1d ago
Related. Others?

China's moonshot launches free AI model Kimi K2 that outperforms GPT4 - https://news.ycombinator.com/item?id=44575309 - July 2025 (3 comments)

Kimi K2 and when "DeepSeek Moments" become normal - https://news.ycombinator.com/item?id=44561565 - July 2025 (2 comments)

Kimi K2 is a state-of-the-art mixture-of-experts (MoE) language model - https://news.ycombinator.com/item?id=44533403 - July 2025 (178 comments)

jtrn•1d ago
The results without the fluff:

Model Architecture * Type: Mixture-of-Experts (MoE) transformer model. * Total Parameters: 1 trillion. * Activated Parameters: 32 billion. * Experts: 384 total experts, with 8 activated per token. * Attention Heads: 64.

Pre-training * Optimizer: A novel optimizer named MuonClip was used. It integrates the Muon optimizer with a QK-Clip mechanism to address training instability. * Dataset: The model was pre-trained on 15.5 trillion tokens. * Training Process: Kimi K2 was trained with zero loss spikes. The initial context window was 4,096 tokens, later extended to 128k tokens using the YaRN method.

Post-training * The model underwent a multi-stage process featuring a large-scale agentic data synthesis pipeline and a joint reinforcement learning (RL) stage. * The RL framework combines verifiable rewards with a self-critique rubric reward mechanism. * A data synthesis pipeline generated tens of thousands of tool-use training examples.

Performance Benchmarks (non-thinking mode) * SWE-bench Verified: 65.8%. * SWE-bench Multilingual: 47.3%. * LiveCodeBench v6: 53.7%. * OJBench: 27.1%. * Tau2-Bench micro-average: 66.1. * ACEBench (en): 76.5. * AIME 2025: 49.5. * GPQA-Diamond: 75.1. * LMSYS Arena Leaderboard (July 17, 2025): Ranked 1st among open-source models and 5th overall.

chisleu•1d ago
It looks like qwen3-coder is going to steal K2's thunder in terms of agentic coding use.
jadbox•1d ago
Maybe so, but currently I like the sound of K2's writing more so than qwen3 (so far in my testing).
swyx•1d ago
(hi i'm OP) kimi k2 was released a while ago with some headlines like muonclip already discussed* but the tech report is new so submitted here. their own highlights are here: https://x.com/Kimi_Moonshot/status/1947520758760313170

we just covered it today on the latent.space paper club if you want to listen along while reading this paper https://youtu.be/VHwZa7lZhK8

definitely see also sebastian raschka's writeup https://t.co/oEt8XzNxik

*background on muon and muonclip https://www.youtube.com/watch?v=fcTNQLebHb0

OutOfHere•1d ago
It has a small context length of just 128K.