Stay informed. Keep your time.
8 stories · ~3 min readEstimated reading time for the entire edition.
Muse Spark Open’s weights have been released. The model is presented as having 816 billion total parameters, 42 billion active parameters, and a 1-million-token context window. The company says it optimized the model to run on two Apple M5 Ultra Mac Studios, and that its license limits commercial use.
Open-weight Muse Spark Open is now available through download and API platforms.
According to an anonymous source who spoke to Axios, the FTC requested information from Anthropic and OpenAI about safety incidents tied to consumer harm, internal safeguards, and marketing. The requests reportedly exceeded 50 pages; the FTC did not provide them to Ars for review.
The FTC requested information from Anthropic and OpenAI about safety incidents and allegations that consumers may have been misled.
Anthropic has given six METR researchers access close to that of company employees, allowing an independent review of its models, safety practices, and risk management. METR may publish its assessments; Anthropic may request removal only of information in narrowly defined, pre-established categories. METR is currently the sole top-tier evaluator in a three-tier system.
Six METR researchers will get employee-level access to Anthropic’s internal tools to examine its safety and risk practices.
Google developed ProtNLM, a deep-learning model that learns evolutionary patterns in amino acid sequences to predict functions from raw protein sequences. Google says it added high-confidence labels to millions of UniProt entries with unknown functions; the text does not explain its results in antimicrobial peptide discovery.
ProtNLM predicts functions from raw protein sequences, adding labels to millions of UniProt entries with unknown functions.
Rosalind announced the Rosalind Academic Plan, offering eligible academic labs sponsored access to GPT-Rosalind and life sciences research tools. Lab administrators at eligible organizations without an Enterprise account can apply; academic users at institutions with an existing Enterprise agreement must request access through their account.
The Rosalind Academic Plan offers eligible academic labs sponsored access to GPT-Rosalind and life sciences research tools.
Ricursive Intelligence is developing a system that lets AI design chips and learn from each iteration to improve subsequent designs. The founders say their earlier system, AlphaChip, produced a circular component layout that surprised engineers and proved successful; the new system optimizes the entire chip.
Ricursive Intelligence is developing a system to automate and accelerate chip design with AI.
Liquid AI released the open-weight LFM2.5-24B-A2B. With 24 billion total parameters and 2 billion active per token, the model performed strongly in the company’s tests of tool use and instruction following, but lagged behind larger models on knowledge-intensive and coding tests.
Liquid AI released the open-weight LFM2.5-24B-A2B model, with 2 billion active parameters, for local agents.
NVIDIA added the open 120-billion-parameter Super VL and two Bolt models for local use to its Nemotron family. According to the company, 74B-A9B is designed to run in 24 GB of memory; the 4 GB E4B model is due out on October 23.
NVIDIA announced the open-weight, 120-billion-parameter Super VL model and two Bolt models for local use.
That’s it for today.The rest of your time is yours.
↑Stories are selected from Hacker News; accessible original articles are summarized and translated with AI. Summaries may contain errors; check the source beneath each story for details. Preparation starts daily at 06:55 Türkiye time (UTC+3), targeting publication before 07:00.
This edition covers articles shared on Hacker News on the date shown above.
Read or listen to a new edition every morning with Kısa for iPhone, save stories for later and keep your streak.
