AI agents, coding assistants, developer frameworks, and APIs.
-
Gemini API Managed Agents: 3.6 Flash, hooks, and more
Google has expanded Managed Agents in the Gemini API, making the new Gemini 3.6 Flash model the default with no code changes required. A new 'environment hooks' feature lets developers run custom scripts before and after tool calls to block, lint, or audit agent actions, alongside new budget caps on total token usage and scheduled, recurring agent runs.
-
Scientific computing in the age of agentic AI
A new OpenAI field report shows scientists using AI coding agents to modernize scientific computing software, including in genomics, accelerating both software development and research discovery. The agents are mainly used to update legacy code and automate repetitive pipeline-building work.
-
MCP startup Runlayer accuses Rippling of stealing its product idea
MCP gateway startup Runlayer has sued HR software company Rippling for trade secret misappropriation, breach of contract, and unfair competition. Runlayer alleges that after nearly a year evaluating its product — during which it shared its roadmap and source code under NDA — Rippling walked away from stalled price negotiations and, according to an internal tip, began building 'a 1-to-1 copy' of Runlayer's gateway.
-
Instacart's CTO says AI made the company stop worrying about tech debt
Instacart CTO Anirban Kundu said at VB Transform 2026 that AI agents now handle 97% of code generation and edits, freeing the company from worrying about technical debt — inactive code simply gets dropped and rebuilt. He said roughly 7,000 automated evaluations run each month and an AI system answers over 8,000 real-time developer queries with about 99.9% accuracy.
-
GM redesigned its engineering workflows around AI agents — and tripled its merged pull requests
GM's VP of autonomous vehicles, Rashed Haq, said at VB Transform 2026 that redesigning engineering workflows around AI agents — not just handing developers a coding assistant — tripled merged pull requests across its autonomous vehicle organization while reducing defects that escape into later development stages. GM's approach was to map each development loop (simulation, road testing, post-deployment monitoring), find its slowest bottleneck, and automate that specific step.
-
Anthropic launches Claude Opus 5, a cheaper AI model for coding, agents and enterprise workflows
Anthropic released Claude Opus 5 on July 24, keeping pricing at $5 per million input tokens and $25 per million output tokens — unchanged from Opus 4.8 — while more than doubling its score on the Frontier-Bench v0.1 agentic coding benchmark (43.3% vs. 18.7%). Opus 5 is now the default model on Claude Max and the strongest model available on Claude Pro.
-
VentureBeat Research: Where enterprise AI agent governance hasn't caught up
VentureBeat Research's five parallel June surveys of 573 companies with 100+ employees found enterprises deployed AI agents ahead of the controls needed to manage them. 71% of respondents said a quarter or fewer of their deployed 'agents' can actually complete multi-step work autonomously — most are chatbots wearing the label — and 69% let multiple agents share a single credential, with those companies suffering a 63.5% security-incident rate versus 40.9% among companies giving each agent its own scoped identity.
-
I tried out OpenAI’s new AI keypad — which will be fun for some coders and slightly mystifying to everyone else
OpenAI unveiled Micro, a $230 specialty keypad built with keyboard designer Work Louder to pair with its Codex coding tool. It has six customizable 'agent' keys on top, six command keys below, and a voice-dictation button, with status lights that shift from white (idle) to blue (thinking) to green (done) to red (error).
-
Agentic coding goes hands-free as OpenAI brings GPT-Live's full duplex voice control to Codex and ChatGPT on the desktop
OpenAI has folded GPT-Live — the full-duplex voice model it launched on July 8 that can listen and speak at the same time — into the ChatGPT desktop app for macOS and Windows, letting developers drive Codex and ChatGPT Work agents by voice alone. Developers can now kick off several tasks at once with a single spoken command — chasing down an auth bug, reviewing a pull request, and generating missing unit tests simultaneously.
-
How Cars24 scales conversations and builds faster with OpenAI
Cars24, an Indian used-car marketplace, says voice and chat agents built on OpenAI's API now handle more than 1 million minutes of customer conversation a month and have brought back 12% of leads that had previously dropped out. Customer-support resolution rates improved by 50%, and turnaround time for key service tasks fell by 80%.
-
One ChatGPT link could smuggle a rogue AI agent into your company
Security firm Zenity discovered "AgentForger," a flaw in OpenAI's Agent Builder that let a single manipulated ChatGPT link spin up an attacker-controlled autonomous agent running with a real employee's permissions and approval checks switched off. The bug exploited a URL parameter to skip approval steps; Zenity reported it via Bugcrowd on June 4, OpenAI confirmed it the next day, and shipped a fix within four days by removing the vulnerable parameter — before the flaw was disclosed publicly.
-
2026 State of AI Agents: Enterprise Insights on Building AI
Databricks released its "2026 State of AI Agents" report, based on anonymized telemetry from more than 20,000 organizations worldwide, including over 60% of the Fortune 500. Multi-agent workflow usage grew 327% between June and October 2025, and AI agents now create 80% of new databases and 97% of database branches on Databricks' serverless Postgres service, Neon. Companies with AI governance frameworks in place shipped 12x more agent projects to production than those without, and 77% of customers now use at least two different LLM families, with 59% using three or more.
-
Introducing OpenAI Presence
OpenAI launched Presence, an enterprise platform for deploying voice and chat AI agents. It bundles guardrails that define policies and pre-approved actions, evaluation tools, and a Codex-powered improvement loop, letting companies hand agents everything from identity verification to real tasks like billing and refunds. OpenAI first tested it on its own customer support phone line, where it matched human-level quality within a few weeks and now resolves about 75% of inbound issues without a human stepping in.
-
Fractal by Plasma AI
Plasma AI has open-sourced 'Fractal,' a tool for building hierarchical agent loops. Each node works iteratively toward a goal in its own git worktree, spinning off separable subtasks into child nodes — so instead of a fixed plan, the tree grows to match the size of the problem. Hard caps on iteration count, depth, number of children, cost, and time bound every loop. In a test run against its own repository, a 187-node tree found 135 issues and fixed 55 of them test-first.