OpenAI scales Trusted Access, adds GPT-5.5-Cyber variant

#1 📝 OpenAI News

Scaling Trusted Access for Cyber with GPT-5.5 and GPT-5.5-Cyber - OpenAI announces scaling Trusted Access capabilities for cyber use cases with GPT-5.5 and a specialized GPT-5.5-Cyber variant, aimed at improving secure, audited access for cybersecurity workflows. The post highlights enhanced safety controls and tooling to support trusted deployments in sensitive cyber environments.

#2 📝 OpenAI News

Advancing voice intelligence with new models in the API - OpenAI introduces new voice-capable models in the API to advance voice intelligence, delivering improvements in speech understanding and generation for developers building voice-enabled applications. The announcement stresses expanded tooling for integration and higher-quality voice interactions.

#3 📝 Simon Willison

llm-gemini 0.31 - Release announcement for llm-gemini 0.31 noting that gemini-3.1-flash-lite is no longer a preview. Links to a previous write-up indicating the model likely hasn't changed since the preview.

#4 𝕏

Philipp Schmid updated the Gemini Interactions API to replace rigid `user`/`model` roles with discrete “steps” (user_input, thought, function_call, tool_call, model_output, etc.), consolidated response_format controls, and added a toggle in the docs.

#5 𝕏

OpenAI launched a Codex Chrome extension that brings debugging browser flows, dashboard checks, research, and CRM updates directly into your browser. It’s live today in the Codex app everywhere except the EU and UK, with support arriving soon.

#6 📝 Claude Code Blog

Collaborate with Claude across Excel, PowerPoint, Word and Outlook - Anthropic announced Claude integrations for Microsoft 365 apps (Excel, PowerPoint, Word, and Outlook) to help teams collaborate and boost productivity within those familiar tools. The post is a product announcement targeted at enterprise users and highlights Claude Enterprise capabilities and agent use cases.

#7 𝕏

xAI launched Grok Voice Think Fast 1.0, a real-world customer-support voice agent that delivers fast, accurate handling of complex workflows even in noisy environments. It scales seamlessly from multi-step troubleshooting to high-volume tool calls.

#8 𝕏

Google DeepMind launched the Gemini-powered AlphaEvolve coding agent, accelerating breakthroughs over the past year in quantum research, biotechnology, logistics, and Google’s AI infrastructure.

#9 𝕏

Anthropic donated Petri, its open-source alignment testing tool, to @meridianlabs_ai for independent development and rolled out a major update enhancing Petri’s adaptability, realism, and depth of tests.

#10 𝕏

Anthropic partnered with Neuronpedia to release NLAs on open models, giving researchers hands-on experience. Try them out at https://www.neuronpedia.org/nla.

#11 𝕏

Cursor introduced /orchestrate, a skill in its SDK that recursively spawns agents to tackle complex tasks. They used it to autoresearch internal skills—cutting token use by 20% and improving evals—and to slash backend cold-start times by 80%.

#12 𝕏

There's An AI For That launched LlamaParse, which converts messy real-world PDFs into clean markdown so LLMs can reason across hundreds of documents at scale.

#13 𝕏

LlamaIndex 🦙 launched a complete browser usage guide for LiteParse, ported by @simonw using Vite hacks and mocking. They plan to streamline the process in future releases.

#14 𝕏

NVIDIA AI breaks down how Perplexity runs on NVIDIA GPUs using the CUTLASS Python stack to optimize AI model inference performance.

#15 𝕏

NVIDIA AI shows how pairing DeepStream with coding agents like Claude Code and reusable Skills can generate full vision AI pipelines from simple natural-language prompts, cutting development cycles from weeks to hours.

#16 𝕏

Harrison Chase dives into how Ramp Labs built Ramp Sheets and their internal AI agent Inspect—unpacking architecture, features, and real-world use cases—on the Max Agency podcast.

#17 𝕏

Andrew Ng launched a short course with CopilotKit co-founder @ataiiam teaching three methods to build chat agents that generate custom UIs—charts, forms, whiteboards—or embed third-party apps on demand.

#18 𝕏

Santiago highlights JetBrains’ new agentic Junie CLI with full bidirectional PyCharm integration, boosting the CLI’s power 10×.

#19 𝕏

Santiago unveiled BACH, a video-generation tool that for the first time keeps character faces—same bone structure, skin tone, and eyes—identical across every single clip. No morphing, no drift—just one consistent character shot after shot.

#20 📝 PromptLayer Blog

Braintrust Alternatives: The Best Prompt Management Platforms in M2026 - A comparison for teams evaluating Braintrust that focuses on operational trade-offs like trace volume, evaluation cost, and speed of shipping changes. The article highlights pricing drivers and practical considerations when choosing a prompt management platform.

#21 📝 Simon Willison

Behind the Scenes Hardening Firefox with Claude Mythos Preview - Mozilla used the Claude Mythos preview to discover and fix hundreds of security vulnerabilities in Firefox, dramatically increasing monthly bug fixes. The write-up details the improved tooling and techniques that made AI-generated reports high-signal and includes notable long-standing bugs and a spike to 423 fixes in April.

#22 in

Peter Yang reports that AI builders are ditching ChatGPT and Claude’s chat for Codex and Claude Code to handle real tasks—editing Google Docs, setting up cron jobs, and shipping features.

#23 in

Udi Menkes notes that leading companies like OpenAI and Apollo are replacing traditional PM titles with “builders”—Deployed Product Managers and Product Builders—to prioritize hands-on, AI-powered development.

#24 𝕏

Aravind Srinivas: Comet has been launched as a standalone browser, letting the Personal Computer AI reliably execute web-based tasks when traditional connectors fall short.

#25 𝕏

clem 🤗 announces the Hugging Face Hub has surpassed 4,000 public RL environments and asks if it’s now the largest platform, inviting suggestions to help it grow further.