OpenAI's frontier models and Codex land on AWS Bedrock, Anthropic quietly files a draft S-1 for a public offering, and JetBrains releases Mellum2 — a 12B MoE model built for code.
OpenAI just made its frontier models — including Codex — available on AWS, meaning builders who already live in Amazon's cloud can now call o3, GPT-4o, and Codex without leaving their existing stack. On the same day, Anthropic quietly filed a draft S-1 with the SEC, signaling it is moving toward a public offering. And JetBrains dropped Mellum2, a 12-billion-parameter Mixture-of-Experts model aimed squarely at code completion.
JetBrains published Mellum2 on Hugging Face on June 2. It is a 12B Mixture-of-Experts architecture, meaning only a fraction of the parameters activate per token — so it runs faster than a dense 12B model on the same hardware. JetBrains built it for code: completions, refactoring, and inline suggestions inside IDEs. The weights are available now on Hugging Face. If you run local models for coding, this is worth a look — it is the first serious MoE coding model from a tooling company that actually ships IDEs.
OpenAI's models are now on AWS Bedrock. That covers frontier models and Codex. For builders already using AWS Lambda, S3, or other Bedrock-connected services, this removes the need to manage a separate OpenAI API key and billing account. You call the models through the same AWS SDK and IAM setup you already use. The announcement was confirmed across OpenAI's blog, Hacker News (213 points), and multiple Reddit communities.
Anthropic filed a confidential draft S-1 with the SEC. A confidential filing is a normal first step before a public IPO — it lets the company work through SEC comments before going public with the document. Anthropic confirmed this on its own news page. The filing does not mean an IPO is imminent, but it does mean Anthropic is formally on that path. For builders who depend on Claude, this matters: a public company faces different pressures on pricing, API access, and product direction than a private one.
OpenAI also announced a Stargate data center build in Michigan. The post frames it as infrastructure for the 'Intelligence Age.' No specific capacity numbers were given in the announcement. Separately, OpenAI published a policy post on its views on AI regulation and political advocacy — worth reading if you follow the policy side, but no product changes attached.
DuckDuckGo is making its no-AI search mode easier to reach as its traffic grows. This is a signal worth watching: a meaningful slice of users actively wants search without AI summaries. If you are building a search-adjacent product, that audience is real and growing.
1. If you have an AWS account, set up a Bedrock API call to OpenAI's o3 or GPT-4o using the AWS SDK. You can wire it into an existing Lambda function without touching your OpenAI billing setup. The AWS docs for Bedrock model access are the starting point.
2. Download Mellum2 from Hugging Face and run it locally with llama.cpp (which just pushed build b9468). Point it at a small codebase and test its inline completion quality against whatever you currently use. JetBrains built this for real IDE use, so it should handle partial-line completions better than general-purpose models.
3. If you are building anything that touches search, add a 'no AI summaries' toggle to your product. DuckDuckGo's traffic bump shows this is not a niche preference — it is a growing user segment you can serve.
Confirmed: OpenAI frontier models and Codex are available on AWS Bedrock as of June 2, 2026 — sourced from OpenAI's own blog and cross-confirmed on Hacker News and Reddit.
Confirmed: Anthropic filed a confidential draft S-1 with the SEC — confirmed by Anthropic's own news post and cross-confirmed on Hacker News (472 points) and Reddit.
Confirmed: JetBrains published Mellum2 (12B MoE) on Hugging Face — sourced from the Hugging Face blog and cross-confirmed via hf_models and Reddit r/LocalLLaMA.
Not independently verified by CBW: We have not tested Mellum2's actual code completion quality or benchmarked it against other local coding models.
Worth noting: Anthropic's S-1 is a confidential draft — the full document is not public yet, so no financials, pricing structure, or IPO timeline can be confirmed from this filing alone.
Source: OpenAI blog — frontier models and Codex on AWS — https://openai.com/index/openai-frontier-models-and-codex-are-now-available-on-aws
Source: Anthropic news — confidential draft S-1 SEC filing — https://www.anthropic.com/news/confidential-draft-s1-sec
Source: Hugging Face blog — Mellum2 launch by JetBrains — https://huggingface.co/blog/JetBrains/mellum2-launch
Source: TechCrunch — DuckDuckGo no-AI search traffic boom — https://techcrunch.com/2026/06/01/duckduckgo-makes-its-no-ai-search-engine-easier-to-access-as-its-traffic-booms/
Source: OpenAI blog — Stargate Michigan data center — https://openai.com/index/stargate-michigan-data-center
Source: GitHub — llama.cpp build b9468 — https://github.com/ggml-org/llama.cpp
Get tomorrow's best AI project in your email. With a guide that works. Free. No spam.