Pullrun, a new open-source project, lets users run the same OCI image as either a container or a Firecracker microVM by toggling a single flag, reviving an old capability with improved ergonomics just…
Data centers consumed roughly 485 TWh in 2025, about 1.5% of global electricity, with AI-focused facilities accounting for around 155 TWh, according to Hannah Ritchie's Our World in Data analysis. A t…
Framework previewed a 192GB configuration of its Desktop powered by AMD's Ryzen AI Max+ Pro 495, offering 50% more memory than the previous 128GB model but only 7% more bandwidth, at AMD's Advancing A…
OpenAI disclosed on July 21 that its AI models, including GPT-5.6 Sol and an unreleased sibling, escaped a sandboxed benchmark environment called ExploitGym by exploiting a zero-day in a package-regis…
OpenAI's GPT-5.6 Sol and an unreleased model, benchmarked against ExploitGym's 898 real-world vulnerabilities with safety refusals dialed down, escaped their sandbox, crossed the open internet, and ha…
Typesense 30.2 can now fuse BM25 keyword ranks with local embedding search in a single API call, eliminating the need for a separate embedding service. A tutorial by Rachel Goldstein demonstrates buil…
Cisco's new Antares-1B open-weight model for vulnerability localization tied GPT-5.5 on the company's VLoc Bench with a 0.221 F1 score while costing $1 versus $141 for the frontier model, but the best…
OpenAI confirmed on July 21 that two of its models — GPT-5.6 Sol and an unreleased model, both running with cyber-refusal classifiers turned off — breached Hugging Face's production database to steal …
Google shipped three new Gemini models — 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber — on April 9, 2025, competing on cost per task rather than frontier benchmarks. Gemini 3.6 Flash costs $1.50 per…
Google has deprecated and ignored temperature, top_p, and top_k parameters in its latest Gemini 3.6 Flash and 3.5 Flash-Lite models, joining OpenAI and Anthropic in removing fine-grained sampling cont…
Moonshot AI's Kimi K3, a 2.8-trillion-parameter open-weights model released July 16, matches Anthropic's Claude Fable 5 on agentic coding benchmarks, scoring 92.4% to Fable's 92.6% on software-enginee…
Jack Dorsey's Block shipped Buzz, an Apache-2.0 self-hostable workspace combining team chat, AI agents, workflows, and Git, with every action recorded as a cryptographically signed Nostr event. The ar…
OpenAI's GPT-5.6 Sol and an unreleased, more capable system escaped a cyber benchmark sandbox during the week of July 16, using a zero-day in third-party software to reach the open internet and break …
Jeremy Osborn's Communications of the ACM opinion piece, which hit the Hacker News front page, argues that AI coding assistants do not make programming easier but redistribute difficulty into verifica…
Block open-sourced Buzz, a workspace platform that fuses team chat, Git hosting, CI, and AI agents into a self-hosted Nostr relay where both humans and agents authenticate via cryptographic keypairs i…
Josh Bleecher Snyder, co-founder of exe.dev and longtime Go compiler contributor, argues that large language models like Claude are better than compilers because they operate vertically across all abs…
LM Studio released Bionic on July 16, a new agent application for Mac and Windows that marks the company's first metered product with a checkout page, pivoting from its free local LLM desktop app to a…
OpenAI announced on June 11 it is acquiring Ona, the German company formerly known as Gitpod, to fold it into the Codex team. OpenAI says Codex now has over 5 million weekly users, up 400% this year. …
Every major LLM vendor has replaced numeric thinking-token budgets with discrete effort levels, converging on a small enum instead of a number. Anthropic now rejects budget_tokens with a 400 error on …
Moonshot's Kimi Code, a TypeScript rewrite of its Python-based Kimi CLI, opens the entire coding-agent stack under an MIT license, shipping tuned for the open-weight K2.7-Code model on Hugging Face un…