Feyn Labs shipped SQRL on July 19, a text-to-SQL model that inspects databases with read-only probes before writing queries, achieving 70.6% execution accuracy on BIRD Dev and edging past Claude Opus …
1Password launched a browser integration on July 16 that lets Claude complete website logins during agentic tasks without the credential ever touching the model, its memory, or Anthropic's systems, us…
Moonshot AI's Kimi K3, a 2.8-trillion-parameter model that topped the Frontend Code Arena leaderboard on day one, releases its full open weights on July 27. The MXFP4 weights require approximately 1.4…
Princeton and Stanford researchers released mini-swe-agent, a coding agent in roughly 100 lines of Python that scores over 74% on SWE-bench Verified, beating Claude Code's scaffolding by six percentag…
Anthropic's Claude Fable 5 model, live for 40 days, leads competitors on SWE-Bench Pro by 11 points with an 80.3% score and costs $10/$50 per million input/output tokens, but its refusal system return…
Seven malicious npm packages impersonating Vite's official @vitejs/* namespace were published between June 29 and July 3, delivering a remote access trojan via blockchain-based command-and-control inf…
A Microsoft Research study of tens of thousands of engineers found that AI coding agents increased merged pull request throughput by 24% over a 16-week period, but the company cancelled most of its Cl…
NVIDIA released Nemotron 3 Embed on July 16, and the 8B model immediately landed at #1 on RTEB with a score of 78.5%, topping Voyage 4 Large, OpenAI text-embedding-3-large, and Cohere embed-v4. The mo…
An unannounced model called Claude Honeycomb EAP briefly appeared in Cursor's production model picker on July 8, with developer @chetaslua posting screenshots showing real metadata and a selectable UI…
Anthropic's July 17 Claude Code update changes /fork to create an independent background session instead of an in-session subagent, introduces /subtask as its replacement for isolated in-session tasks…
Georgia Tech's Systems Software and Security Lab has confirmed 74 CVEs directly attributed to AI-generated code in 2026, with the actual number estimated at 400 to 700 cases in the open-source ecosyst…
GitHub Copilot Vision is generally available as of July 1, 2026, allowing all Copilot subscribers, including free-tier users, to attach images, screenshots, UI mockups, and PDFs into the chat panel fo…
Google Cloud Run Sandboxes, now in public preview as of July 10, lets developers run arbitrary code from AI models inside existing Cloud Run instances with no additional infrastructure or cost, enforc…
VS Code 1.129, released July 15 by Microsoft, introduces the Agent Host, which moves AI coding agents into a dedicated background process, preventing agent failures from freezing the editor and enabli…
Cloudflare has launched Precursor, a session-wide behavioral detection engine that monitors mouse movements, typing cadence, and scroll rhythm continuously to identify bots, replacing traditional poin…
Stack Overflow's monthly question volume has collapsed 99.41% from a 2014 peak of 207,000 to roughly 1,226 today, with the steep decline beginning in November 2022 coinciding with ChatGPT's launch, ac…
GoPro faces a going-concern warning from its auditor PricewaterhouseCoopers, which on June 1, 2026 expressed substantial doubt about the company's ability to survive another 12 months. The cause is a …
XAI shipped Grok 4.5 on July 8, priced at $2 per million input tokens and $6 per million output tokens, positioning it as "Opus-class at half the cost" but ranking fourth overall on the intelligence l…
Anthropic's Claude Code dynamic workflows, now generally available for all paid plans, enable up to 1,000 parallel subagents with adversarial verification, as demonstrated by Jarred Sumner porting 750…
VS Code 1.128, released July 8, introduces multi-chat agent sessions that allow parallel conversations with separate models and histories within a single workspace, and makes Copilot Vision generally …