OpenAI’s GPT-5.5 and Codex now on AWS Bedrock

#1 📝 OpenAI News

OpenAI frontier models and Codex are now available on AWS - OpenAI made its frontier models (including GPT‑5.5) and Codex generally available on AWS via Amazon Bedrock on June 1, 2026, enabling enterprises to run those models in Commercial and GovCloud regions using AWS-native security, procurement, billing, and governance workflows. Codex is claimed to be used by more than 5 million people weekly, and OpenAI says future offerings like Daybreak — cyber models and Codex Security for secure code review, threat modeling, patch validation, dependency risk analysis, detection, and remediation guidance — will be brought to AWS.

#2 𝕏

OpenAI frontier models and Codex are now generally available on AWS Bedrock, letting enterprises build on them through their existing security, compliance, and governance workflows. Future expansions will bring capabilities like the Daybreak cybersecurity offering.

#3 𝕏

Anthropic confidentially filed a draft S-1 registration statement with the U.S. Securities and Exchange Commission, paving the way to pursue an initial public offering pending SEC review.

#4 𝕏

Philipp Schmid launched Managed Agents in the Gemini API, allowing users to spin up autonomous AI agents that reason, write and run code, and manage files inside a hosted Linux sandbox with just one API call.

#5 𝕏

Qwen launched Qwen3.7-Plus, a single multimodal agent that unifies vision and language with GUI/CLI support for coding, visual reasoning, grounding, and search-augmented QA. It’s now available via API on Alibaba Cloud Model Studio.

#6 𝕏

NVIDIA AI introduced Factory Operations Blueprint (FOX), a reference design for AI-powered factory manager agents that monitor operations, analyze real-time data, and coordinate specialized AI agents to resolve issues at scale.

#7 𝕏

xAI launched Composer 2.5 inside Grok Build, a fast, highly intelligent model optimized for long-running tasks and following complex instructions.

#8 𝕏

Logan Kilpatrick shipped integrated Gmail, Drive, and Sheets connectors in Google AI Studio, letting you build and test apps without leaving the platform. You can now add testers right inside AI Studio, with full public sharing coming soon.

#9 𝕏

Harrison Chase (@LangChain) shares a behind-the-scenes look at how Rippling built RipplingAI, detailing their tech stack, modular architecture and integration approach to embed generative AI across HR/IT workflows.

#10 𝕏

Santiago launched “The Grid,” an OpenAI-compatible API marketplace that routes requests to the cheapest model in your chosen quality tier (standard, prime, max) via a one-line code change and audits suppliers to ensure quality. New accounts get 200 M free tokens.

#11 𝕏

Santiago launched Ingestr, a Python CLI tool supporting four incremental ingestion strategies—replace, append, merge, and delete+insert. Install via `pip install ingestr` and move data with `ingestr ingest --source-uri --dest-uri` (github.com/bruin-data/ingestr).

#12 𝕏

Peter Yang shares six key insights from @Shpigford on solo AI-agent product building: ship despite fear and charge users early, use Git worktrees for code management, formalize model reviews, build a /learnings skill for ongoing improvement, and leverage domain expertise to s...

#13 𝕏

Aravind Srinivas (CEO, Perplexity) says Perplexity is shifting search from a simple web-fetch tool to a code generation engine, enabling agents to compose multi-step primitives more naturally.

#14 𝕏

clem 🤗 cloned a 68 TB dataset into a private HF training bucket in under a minute—despite only having a 4 TB local disk—thanks to HF infra optimizations and xet dedup!

#15 𝕏

Julien Chaumond dropped a new docs page on the Hugging Face Hub detailing how to render Agent Traces, enabling clear visualization of agent workflows.

#16 𝕏

Guillermo Rauch announces that MiniMax M3 now tops Next.js agent evaluations among open models—placing just behind Opus and GPT-5 while costing 10× less (and currently 20× cheaper on ▲ AI Gateway).

#17 📝 PromptLayer Blog

How to write an LLM prompt spec - An LLM prompt spec is defined as an engineering contract that must specify a prompt’s purpose, inputs, outputs, constraints, evaluation criteria, ownership, and failure modes, with concrete sections such as name/owner/feature/model/fallback/runtime, a tight task definition example (e.g., classify support tickets into exactly one category: billing, technical_support, account_access, abuse, or other and return only JSON), an inputs table (ticket_subject max 200 characters, ticket_body truncate after 4,000 tokens, retrieved_policy_snippets max 5), and explicit context-budget targets (system/dev instructions <800 tokens, user content <4,000 tokens, retrieved snippets <2,000 tokens, examples <1,200 tokens, total target <8,000 tokens). It also mandates separating instructions, policy, and user data to mitigate prompt injection and provides an explicit JSON output schema example (required properties: category enum, confidence number between 0 and 1, rationale string maxLength 300) plus valid and invalid output examples.

#18 📝 Simon Willison

Pasted File Editor - A prototype Pasted File Editor that detects large pasted text and turns it into a file attachment, with support for opening files and showing image thumbnails. Built as a prototype using Codex desktop.

#19 ▶️

Building a Agentic AI Trading Heartbeat That Works

All About AI

#20 in

Greg Isenberg unveiled GPT Realtime 2.

#21 📝 Simon Willison

Hackers Simply Asked Meta AI to Give Them Access to High-Profile Instagram Accounts. It Worked - Reports indicate attackers used Meta's AI support bot to perform account recovery and link attacker-controlled emails to high-profile Instagram accounts, enabling takeovers. Simon warns against wiring support bots into account recovery flows that allow one-shot takeovers.

#22 𝕏

Andrew Ng highlights the rise of AI Forward Deployed Engineers—client-embedded specialists customizing and tuning agentic workflows—and predicts that, despite OpenAI and Anthropic expanding FDE teams, AI Engineer roles will far outnumber FDE positions.

#23 𝕏

Google AI used Bard’s new multimodal model, Vertex AI Vision real-time object detection, and on-device translation to power interactive demos, signage, and attendee experiences at I/O 2026.

#24 𝕏

Cursor is boosting usage limits for all Teams users and, riding on the Ultra plan’s success, is launching a Premium team seat that delivers 5× the usage for only 3× the cost.

#25 𝕏

Lenny Rachitsky distills Benedict Evans’s 10 AI takeaways: we’re at a ’97-PC style inflection, facing risks like the Jevons paradox alongside emerging distribution moats and model pricing power.