OpenAI’s GPT-5 debuts with enhanced features across sectors, Claude Opus 4.1 elevates coding prowess for developers, and Google DeepMind presents Genie 3 for interactive environments
GPT‑5 is smarter across the board, providing more useful responses across math, science, finance, law, and more. It also produces high-quality code, generates front-end UI with minimal prompting, and shows improvements to personality, steerability, and executing long chains of tool calls.
Anthropic released Claude Opus 4.1, an upgrade with state-of-the-art performance in coding, reasoning, and agentic tasks. Available now for paid users and via the API, it offers notable gains for developers, with more updates coming soon.
OpenAI releases gpt-oss-120b and gpt-oss-20b: Apache-2.0 open-weight models with strong tool use and 128k context. 120b nears o4-mini and runs on one 80GB GPU; 20b matches o3-mini and fits 16GB devices. Weights (MXFP4), tokenizer, and tools ship with a safety-vetted model card.
Google DeepMind unveils Genie 3, a real-time world model that generates interactive 720p environments at 24 fps from text prompts, keeping them consistent for minutes. It adds promptable world events, supports embodied-agent research, and launches as a limited research preview.
xAI’s Grok Imagine rolls out on X’s iOS for SuperGrok and Premium+ users, generating images and 15-sec videos from prompts. A “spicy mode” allows NSFW with moderation and celebrity limits; results feel uncanny, but the UX is fast and slick.
At a Reddit AMA, Sam Altman said GPT-5 seemed “dumber” because the autoswitcher failed at launch. He promised fixes, clearer model transparency, doubled Plus rate limits, is considering restoring 4o for Plus, and called the benchmark chart a “mega screwup.”
X will place ads inside Grok’s answers, collapsing the line between utility and promotion. The piece defines AI Neutrality’s pillars (consent, separation, pluralism, transparency, sensitive-context limits), warns of manipulation and privacy risks, and urges user-controlled fetch-agent models.
OpenAI launches GPT-5 days after its open models and despite Altman calling it “the best,” it only slightly beats rivals on some benchmarks. That said, it's pricing ($1.25/M input, $10/M output, $0.125/M cached) pressures Google and undercuts Anthropic.
AI coding agent, built for the terminal.
Cursor Agent now runs via CLI/headless in any environment, alongside Neovim, JetBrains, or other IDEs and can run multiple agents in parallel. It works with any model in your subscription, however it’s still in beta with broad file/command access, so use in trusted environments.
Automated security reviews land in Claude Code via a /security-review command and GitHub Action. It scans codebases for SQLi, XSS, auth, insecure data, and dependency risks, comments inline on PRs, suggests fixes, and has already caught remote code execution and SSRF.
Gemini CLI GitHub Actions is a no-cost, powerful AI coding teammate for your repository. It acts both as an autonomous agent for critical routine coding tasks, and an on-demand collaborator you can quickly delegate work to.
Jules, the asynchronous coding agent, launches publicly with Gemini 2.5 Pro. The beta yielded over 140k code improvements and added setup reuse, GitHub issues, and multimodal. New tiers rolling out include: Intro, Pro (5× limits), Ultra (20×).
It’s now much better at challenging and long-running tasks, especially in large codebases.
Rolling out to Max, Team, and Enterprise plans today, with other plans coming soon.
After submitting a ticket, Cursor harvested and pulled all credentials from the dev machine.
Start your dev server, run tests, or build your project without blocking your workflow.
AI models have unstable personalities. Researchers can now identify "persona vectors"—neural network patterns that control specific traits like evil or sycophancy. This allows for monitoring and mitigating undesirable behaviors to keep AI aligned and safe.
A curated weekly roundup of AI news, tools, and breakthroughs to stay ahead of the curve in a fast-moving industry.
Curated by devs, for devs. Helping 400+ developers stay ahead of the AI curve.