PickAIModel Brief - Issue 007

Issue 007 — June 8, 2026 Independent guide to the AI industry, useful tools, and practical model choices.

The ten days from May 29 to June 8 were not about another leaderboard shuffle. Anthropic started its IPO process at a $965 billion private valuation, Microsoft used Build to show why it no longer wants to depend entirely on OpenAI, Congress released a federal AI bill that reopens the state-preemption fight, and OpenAI changed how ChatGPT carries memory across conversations. These are platform stories. They affect pricing, trust, data governance, procurement risk, and the tools people actually use.

The Story That Sets the Issue

Anthropic Filed Its S-1. The Private AI Era Is Starting to End.

Anthropic has started the IPO process before OpenAI. On June 1, Anthropic said it had confidentially submitted a draft Form S-1 registration statement to the SEC, four days after announcing a $65 billion Series H at a $965 billion post-money valuation. The round was led by Altimeter Capital, Dragoneer, Greenoaks, and Sequoia Capital. That valuation puts Anthropic within touching distance of the trillion-dollar mark before public investors have seen the full risk disclosures, revenue concentration, infrastructure commitments, or customer mix.

The public S-1 is not available yet, so the financial picture is still partial. Reported private figures, however, explain why the filing matters. Reuters Breakingviews reported that Anthropic generated $4.8 billion in Q1 2026 revenue and projected $10.9 billion for Q2. If that Q2 projection holds, it implies an annualised run rate approaching $44 billion, up from a reported roughly $9 billion annualised at the end of 2025. Even by AI-boom standards, that is an extraordinary acceleration.

Why it matters: The AI labs are moving from private-market storytelling into public-market disclosure. Buyers negotiating enterprise AI contracts should expect the pricing conversation to change once investors can see margins, infrastructure obligations, and customer concentration in public filings. The best contract terms are often available before the earnings-call discipline begins.

The Stories That Matter

Microsoft Used Build 2026 to Loosen OpenAI’s Grip on Its AI Stack

Microsoft did not leave OpenAI. It did something more interesting: it made dependence optional. At Build 2026, Microsoft introduced the MAI model family, led by MAI-Thinking-1, a 35-billion-active-parameter mixture-of-experts model with a 256,000-token context window, and MAI-Code-1-Flash, a 5-billion-active-parameter coding model rolling out to GitHub Copilot individual users in VS Code. Microsoft AI says MAI-Code-1-Flash was designed for Copilot workflows and will appear through the Auto picker or directly in the model picker as rollout progresses.

That distinction matters. Copilot and Azure still support OpenAI models, and Microsoft remains deeply tied to OpenAI through infrastructure and investment. But Microsoft now has first-party models it can tune, price, package, and govern on its own terms. For large customers, that changes the choice from “Microsoft plus OpenAI” to “Microsoft, OpenAI, open models, or a mixture.”

Why it matters: Enterprise AI buying is becoming a portfolio decision. If you are standardising on Microsoft tooling, the question is no longer just which frontier model is best. It is which model has the right cost, provenance, data-handling terms, latency, and integration path for the workflow.

Congress Released Its Biggest AI Bill Yet. The Fight Is Over State Preemption.

The Great American AI Act is not law. It is a warning shot. On June 4, Representatives Jay Obernolte and Lori Trahan released a discussion draft of a federal AI framework, with Reuters reporting that the draft would bar states from regulating AI model development while still allowing states to regulate AI use. Axios described it as a 269-page draft covering whistleblower protections, third-party evaluations, fraud enforcement, AI literacy, and federal oversight.

The political fault line is clear. Tech groups want one national rulebook. Consumer and labor groups worry that federal preemption would weaken the state-level protections that have emerged while Congress has struggled to legislate. The AFL-CIO, AFT, and Association of Flight Attendants issued a joint rejection of the draft, while Public Citizen said the proposal would strip states of authority to protect consumers, workers, and children.

Why it matters: Builders should not assume the US regulatory map will stay fragmented forever, but they also should not treat federal preemption as inevitable. If the bill advances, compliance architecture for model developers could shift toward federal obligations. If it fails, the state-by-state patchwork continues to expand.

OpenAI Changed How ChatGPT Carries Memory Across Conversations

ChatGPT memory is becoming less like a notepad and more like a continuity layer. On June 4, OpenAI’s release notes said it had upgraded memory so ChatGPT can better keep context up to date, reduce stale or contradictory saved memories, and better understand a user’s preferences, goals, and ongoing work. OpenAI’s Memory FAQ says the new memory system is starting with Plus and Pro users in the US and will expand to more plans and countries over the coming weeks.

That is useful. It is also a governance issue. A system that remembers your preferred writing style, business context, projects, decision patterns, and recurring constraints can produce better work with less prompting. The same continuity creates a persistent context layer that users and companies need to understand, review, and manage. OpenAI says users can review memories that may be used for personalization, and Temporary Chat remains available for conversations that should not draw on or update memory.

Why it matters: For individual users, better memory means less repetitive prompting. For teams, it means memory settings, account type, retention policy, and workspace controls are no longer secondary details. They are part of the risk model for using ChatGPT in professional work.

Technology Shift

Microsoft’s Majorana 2 Claim Is a Hardware Story With an Asterisk

Microsoft’s most interesting Build announcement may not have been a model. It may have been a chip. Microsoft said its Majorana 2 quantum chip uses a new materials stack that delivers a 1,000-fold reliability improvement over the prior generation of qubits, with a mean qubit lifetime of 20 seconds and some instances lasting as long as one minute. Reuters reported that Microsoft says the lead-based design produced a 1,000-fold performance boost in some areas and helped the company target commercially useful quantum computers by 2029.

The asterisk is important. Microsoft’s topological-qubit programme has faced years of scrutiny, and Nature reported that researchers remain sceptical about whether the company has provided enough transparent, reproducible evidence for the underlying claims. Microsoft says it has shared sufficient data with agencies such as DARPA and stands by the physics.

Practical significance: if validated, longer-lived qubits would matter for optimisation, materials science, cryptography, and eventually parts of AI infrastructure. For now, treat Majorana 2 as a serious claim from a serious lab, not as settled commercial capability.

Tool of the Issue

GitHub Copilot with MAI-Code-1-Flash Microsoft’s first-party coding model is rolling out to Copilot individual users in VS Code.

MAI-Code-1-Flash is now rolling out to GitHub Copilot individual users in Visual Studio Code through the Auto picker and model picker. Microsoft describes it as a lightweight, agentic coding model built for Copilot and VS Code, with no additional setup required as the rollout reaches eligible users. GitHub’s changelog also says MAI-Code-1-Flash is rolling out in GitHub Copilot starting with VS Code.

Best for: Developers already using GitHub Copilot who want to test Microsoft’s first-party coding model inside their normal editor workflow.

Caution: Microsoft’s performance claims are still Microsoft’s claims. The model is worth testing on your own repositories before drawing conclusions from launch materials.

Why this issue: It costs nothing extra if you are already an eligible Copilot user, and it is the first practical test of whether Microsoft’s own coding models can earn usage inside Microsoft’s most important developer product.

GitHub Copilot | Non-affiliate link

The Number That Changes the Picture

80%

That is the share of code merged into Anthropic’s codebase that was authored by Claude as of May 2026, according to Anthropic’s own Institute essay, “When AI builds itself”. Anthropic says that before Claude Code launched in research preview in February 2025, the figure was in the low single digits. The number matters because it is not a survey answer or a demo claim. It is a production-engineering signal from a frontier AI lab using its own coding agent at scale. At Anthropic, AI-authored production code is no longer a future scenario; it is part of normal engineering workflow.

Accidental Honesty

At Build, Microsoft’s message was that every company should move from consuming frontier models to participating at the frontier. It was aimed at customers. It also described Microsoft’s own shift. For five years, Microsoft funded, hosted, packaged, and sold access to frontier models it did not fully own. The MAI launch is the point where that dependence starts becoming optional.

Watch List

• Anthropic’s public S-1: the confidential filing is in. The public version, when it appears, should reveal customer concentration, infrastructure obligations, risk factors, and the real economics behind Claude’s growth.
• The Great American AI Act: the preemption fight will decide whether US AI governance moves toward one federal framework or a growing patchwork of state obligations.
• Microsoft’s MAI adoption: the important signal is not a launch chart. It is whether Copilot and Azure customers actually choose Microsoft’s own models when OpenAI and Anthropic models remain available.

If someone forwarded this to you, subscribe at pickaimodel.com/newsletter. Have a question for a future issue? Reply to this email — we read everything.

Disclosures

Editorial independence. PickAIModel.com produces independent editorial content. Model rankings, quality scores, and value scores are determined by our published methodology and are not influenced by commercial relationships with any AI vendor. No company can pay for ranking position, score changes, inclusion in rankings, or favourable treatment in our methodology outputs.

AI-assisted content disclosure. Portions of the editorial summaries and commentary in this newsletter may be drafted with the assistance of AI language models and reviewed by the PickAIModel editorial team. Benchmark scores, pricing fields stored in PickAIModel’s own system, and ranking data are not invented by the model and are not editorially altered to favour advertisers or affiliates. Routine leaderboard comparisons are handled on PickAIModel.com rather than repeated in this newsletter unless a benchmark event is unusually consequential.

Affiliate disclosure. This newsletter may contain affiliate links. If you click a qualifying link and make a purchase, PickAIModel.com may earn a commission at no additional cost to you. Affiliate relationships do not influence our rankings, scores, or methodology outputs. The Tool of the Issue link this edition is a non-affiliate link.

Not financial or legal advice. Nothing in this newsletter constitutes financial, investment, or legal advice. References to funding rounds, valuations, IPO timelines, pricing, contracts, or company strategy are provided for informational purposes only. Make decisions based on your own judgment and, where appropriate, qualified professional advice.

Accuracy and currency. AI model pricing, capabilities, availability, and company claims change frequently. While we aim to be accurate at the time of publication, information may become outdated or be revised after publication. Verify critical purchasing details directly with the relevant vendor before acting.

To unsubscribe, use the unsubscribe link in your email footer. Copyright © 2026 PickAIModel.com. All rights reserved. NeuroSync AI Dynamics Pty Ltd.