Liquid AI open-weights two CPU-fast 8K text encoders
LFM2.5-Encoder-230M and 350M bring an 8,192-token window to BERT-class models and cut CPU latency roughly threefold at full length.
July 28, 2026
All briefings and quick hits published that day.
Sangfor says its GLM-5.2 agent hits 86.3% on CyberGym
The Chinese security vendor reports its agent reproduced 1,301 of 1,507 vulnerabilities on CyberGym, claiming fourth place globally atop open GLM-5.2 weights.
AI chip selloff hands Kospi its worst two-day drop ever
Seoul's benchmark fell about 6% on Wednesday after a 10.84% Tuesday collapse, as SK Hynix missed estimates and China's first domestic DUV tools entered service.
OpenAI open-sources its Codex Security scanning CLI
The Apache-2.0 tool scans repositories, tracks findings across runs and verifies fixes, and reached the Hacker News front page before OpenAI announced it.
Cyera to buy Oasis Security for $1B to police AI agents
The data-security firm signed a letter of intent for the non-human identity startup, its third acquisition this year, as enterprises scramble to govern agent access.
- ⚡ Quick hits
- Policy The UK's CMA opened an investigation into whether Microsoft misled Microsoft 365 Personal and Family customers when it bundled Copilot and rolled them onto pricier plans. GOV.UK (CMA) ↗
- Agents Fireworks AI launched Nexus, a drop-in router that sends routine coding requests to open-weight models, reporting 3–5x cost cuts and 33% lower cost per merged PR. MarkTechPost ↗
- Safety JFrog detailed how OpenAI's agent used a zero-day in its app to breach Hugging Face, framing the exploited disclosure as a security success story. Ars Technica ↗
- Funding Bot-detection startup Spur Intelligence raised $200 million from Insight Partners as agentic AI traffic pushes bots past humans on the internet. TechCrunch ↗
- Agents MCP gateway startup Runlayer sued Rippling, alleging it cloned Runlayer's product after a year-long NDA trial in which roadmap and source code were shared. TechCrunch ↗
OpenAI ships GPT-Transcribe and GPT-Live-Transcribe APIs
Two new speech-to-text models replace Whisper as OpenAI's recommended default, with steerable context inputs and pricing from $0.0045 per minute.
- ⚡ Quick hits
- Research Ars Technica reported Google's own workplace data shows employees using AI to augment tasks rather than automating their own jobs away. Ars Technica ↗
More than 1,300 AI staff ask Washington to pace automated AI R&D
A letter signed by researchers at Anthropic, OpenAI, Google DeepMind, Meta, Thinking Machines and Safe Superintelligence asks the US to lead an international effort to build tools for slowing frontier AI. The signature count is still climbing — 1,324 as of 1 August.
- ⚡ Quick hits
- Models Perplexity added Moonshot's Kimi K3 to Perplexity and Perplexity Computer for Pro and Max subscribers, hosted exclusively on US-based servers. X @perplexity_ai ↗
- AIGC Ideogram said it is partnering with AMD to serve its open-weight Ideogram 4.0 image model at scale on Instinct accelerators. X @ideogram_ai ↗
- AIGC PixVerse showcased a girl-group music video shot as one continuous orbiting take, with the camera circling and revealing each member without cuts. X @PixVerse_ ↗
- Models LMArena said xAI's Grok 4.6 arrives August 7 and will enter the arena leaderboards next week. X @arena ↗
Claude Mythos finds new attacks on HAWK and 7-round AES
Anthropic says its frontier model autonomously produced two publishable cryptanalysis results, cutting HAWK-256's effective security and speeding an AES attack 200-800x.
- ⚡ Quick hits
- Research OpenAI published a report on agentic scientific computing, using eight industry and academic case studies to argue coding agents shift researchers toward verification and stewardship. OpenAI ↗
- Open Source Ai2 detailed its OlmoEarth Platform, which ran a North America wildfire-risk job on 19,600 CPUs and 994 GPUs for a 155x speedup over serial inference. Hugging Face ↗
- Research LMArena's Code Arena added fullstack web development rankings, scoring models on multi-step reasoning and tool use across complete app-building tasks. X @arena ↗
- Research World Labs presented real-to-sim-to-real as a scalable engine that turns one physical robot task into many controllable simulated worlds for policy training. X @theworldlabs ↗
Amazon freezes Nova line, bets on new Frontier model team
Business Insider reports Amazon has halted development on Nova Premier, Omni, Reel and Canvas, shifting resources to a research group under Pieter Abbeel.
Gemini API managed agents add 3.6 Flash and sandbox hooks
Google's hosted agent runtime switches its default to Gemini 3.6 Flash and adds pre/post tool-call hooks, token ceilings, cron triggers and a free tier.
- ⚡ Quick hits
- Policy PJM, the largest US grid, will curtail power to data centers of 50MW or more during shortages from June 2027, with advance notice and compensation. TechCrunch ↗
- Agents ElevenLabs extended ElevenAgents to SMS, Telegram, email and ticketing, with per-channel response controls so one agent keeps the same knowledge, tools and guardrails everywhere. X @ElevenLabs ↗
- Agents Hugging Face published Training Agents 3, a hands-on unit on training local open-weight agents with reinforcement learning. X @huggingface ↗
- Research LMArena said Meta's Muse Spark 1.1 also moved the Vision Pareto frontier, scoring 1283 on its vision leaderboard. X @arena ↗
- Open Source Liquid AI open-weighted LFM2.5-Encoders, 230M and 350M encoders with 8,192-token context that run about 3.7x faster than ModernBERT-base on CPU. Hugging Face ↗
- AIGC Runway announced that ByteDance's Seedance 2.5 video model is coming soon to its platform, adding a rival lab's generator to Runway's tool lineup. X @runwayml ↗
- Funding Fish Audio raised a $52M seed led by Coreline Ventures and Capital Today for voice models with 15,000+ controls, citing 8M users and $21M ARR. TechCrunch ↗
- Industry Richard Socher's Recursive Superintelligence signed a $410 million multiyear compute deal with AWS, with no investment component and co-developed infrastructure. TechCrunch ↗
- Industry Nvidia is investing in Ilya Sutskever's Safe Superintelligence, moving the ~$30B-valued lab off Google TPUs onto Vera Rubin GPUs for a claimed tenfold compute boost. The Decoder ↗
Perplexity's Personal Computer agent lands on Windows
The desktop agent reads local files, drives Office apps and routes subtasks across 20-plus frontier models, gated behind Perplexity's $200-a-month Max tiers.
- ⚡ Quick hits
- Policy Anthropic CEO Dario Amodei reaffirmed his warnings about open-weight model risks while insisting he never called for banning open models. The Decoder ↗