b10278
build : remove GGML_METAL_USE_BF16 from all build scripts (#26604) Website: https://llama.app macOS/iOS: macOS Apple Silicon (arm64) macOS Apple Silicon (arm64, KleidiAI enabled) DISABLED macOS Intel (x64)…
Full feed
Feed 28,181 posts
Every story across every category, newest first. Each card links to the original publisher; daily-brief posts open as editorial pages.
An AI agent went rogue during UK safety tests, creating fake identities and launching social engineering attacks unprompted
In a security test by the British AI Safety Institute, an AI agent went rogue on the open internet without being told to. It created fake…
Chinese insistence that the US is not allowed to "generalize" or "overstretch" the concept of national security is a de facto attack on hegemony. The …
Chinese insistence that the US is not allowed to "generalize" or "overstretch" the concept of national security is a de facto attack on hegemony. The US…
Introducing GPT-Daybreak to accelerate defenders
Frontier cyber models for defenders, from broad defensive work to advanced security research.
Putting frontier cyber models in more trusted hands
Approved Daybreak partners can use OpenAI’s frontier cyber models to deliver authorized, governed cybersecurity services to customers.
Alibaba, DeepSeek push China’s AI model race towards lower costs
Alibaba has launched Qwen3.8-Max, its largest AI model to date, as DeepSeek’s latest V4-Flash model draws attention for inference pricing that is lower than several competing…
Making difficulty curves in games
Article URL: http://www.davetech.co.uk/difficultycurves Comments URL: https://news.ycombinator.com/item?id=49180649 Points: 100 # Comments: 37
RT SpaceX: Falcon 9 launches from pad 40 in Florida
RT SpaceXFalcon 9 launches from pad 40 in Florida
The software group Palantir paid just £2M in corporation tax in the UK in 2024
Article URL: https://www.theguardian.com/technology/2026/aug/05/palantir-corporation-tax-uk-2024-accounting-practices Comments URL: https://news.ycombinator.com/item?id=49180535 Points: 6 # Comments: 2
DeepSeek-V4-Flash-Preview vs V4-Flash-0731, same problem 0731 can solve it, Preview could not Good Guy Grug. "need exact"
DeepSeek-V4-Flash-Preview vs V4-Flash-0731, same problem0731 can solve it, Preview could notGood Guy Grug. "need exact"kache: Deepseek flash is genuinely really good. I haven't experienced any reward…
Helsinki Hacker News Meetup
Article URL: https://calpaterson.com/helsinki-hn.html Comments URL: https://news.ycombinator.com/item?id=49180407 Points: 78 # Comments: 34
RT antirez: Now part of the DwarfStar README
RT antirezNow part of the DwarfStar README
Decoding Children's Gait Behavior
We introduce a new problem domain for human action recognition: the fine-grained analysis of children's gait behaviors from standard RGB video. We specifically target the ambulatory…
MiniWorld: Democratizing the Training of Video World Models from Scratch
Video world models predict future observations conditioned on historical observations and control signals, enabling long-horizon generation through autoregressive state transitions. Unlike conventional video generation models that…
Fuck, another round of the trade war
Fuck, another round of the trade warTeo: Announcement of the Ministry of Commerce No. 34 of 2026 announced the strengthening of export control of dual-use items…
I like the pitch It's very Purist. "no I am not satisfied with a system prompt and Memories. I need mine"
I like the pitchIt's very Purist. "no I am not satisfied with a system prompt and Memories. I need mine"will depue: gwern is a generational talent,…
Will be glorious if they stick the landing
Will be glorious if they stick the landingᯅ̈?!: Zhuque-3 Y2 is scheduled to launch during 07:37–08:14 on August 11, 2026 for first-stage recovery mission. Success is…
AI Morning Brief — 5 August 2026
Overnight in AI: Anthropic commits $10 billion of compute to a brand-new cloud startup while UK evaluators publish cybersec findings on Mythos 5 and GPT-5.6.
Building a Fully Local PDF Read-Aloud & PDF-to-Audiobook Desktop App with Kokoro 82M, Qwen, and llama.cpp
Hey everyone, I’ve been building Speechfony - a desktop app for reading PDFs (and EPUBs) with offline text-to-speech. Open a document, listen sentence-by-sentence with highlighting, or…
NVIDIA Releases Alpamayo 2 Super: A 34B Open Vision-Language-Action Model for Robotaxis and Autonomous Driving Under OpenMDW-1.1
NVIDIA released Alpamayo 2 Super, a 34B vision-language-action model for autonomous driving, under OpenMDW-1.1 — a permissive license covering fine-tuning, derivatives and commercial redistribution. It pairs…
ContinualSkillBench: Can LLM Agents Truly Evolve Their Capabilities?
Modern agent frameworks equip large language models with external skill libraries to solve complex tasks. However, it remains unclear whether these systems can effectively evolve their…
RT SpaceX: Falcon 9’s first stage has landed on the A Shortfall of Gravitas droneship
RT SpaceXFalcon 9’s first stage has landed on the A Shortfall of Gravitas droneship
RT xlr8harder: It's coming
RT xlr8harderIt's comingSamuel Hammond 🦉: A snapshot of some of the unsanctioned actions Mythos took while attempting to poison an open-source project
Qwen3-TTS voice cloning is now in mainline llama.cpp — the old demo finally became real support
People may remember the Qwen3-TTS llama.cpp demo from a few months ago. That PR said it probably wouldn’t be merged because llama.cpp was missing some of…