In partnership with
Today on FindMeAI:
For two years, the AI vendor stack looked frozen — OpenAI on top, Anthropic chasing, Microsoft riding both. This week, all three positions moved at once.
→ Anthropic passed OpenAI — $965B valuation, $47B run rate, and Opus 4.8 tops every coding benchmark
→ Microsoft Build dropped homegrown MAI models — and a direct shot at Claude Code's dominance
→ OpenAI shipped Codex defense — Computer Use on Windows + GPT-5.5 quality fixes
→ WWDC keynote tomorrow — Siri 2.0, multi-model AI, what to position for before Monday
Anthropic just passed OpenAI — and shipped the model that earned it.
On May 28, Anthropic closed a $65B Series H at a $965B post-money valuation, passing OpenAI's $852B for the first time. Run-rate revenue crossed $47B earlier in May. Same day: Claude Opus 4.8 launched — 41 days after 4.7, Anthropic's fastest cadence ever, priced identically to the prior Opus.
Opus 4.8 tops the benchmarks Anthropic chose to publish:
→ Beats GPT-5.5 and Gemini 3.1 Pro on agentic coding
→ Beats both on financial analysis
→ Beats both on computer use
The headline feature is what shipped underneath: Dynamic Workflows. Opus 4.8 paired with Claude Code can spin up 1,000 parallel subagents working on one task — built for repo-scale migrations end-to-end. Currently research preview.
Anthropic | OpenAI | |
|---|---|---|
Valuation (May 28) | $965B | $852B |
Run rate | $47B | ~$25B |
Q2 bottom line | +$559M projected profit | −$14B operating loss projected |
OpenAI co-founder Andrej Karpathy joined Anthropic's pre-training team the week before. Anthropic also teased its next model, "Mythos," for public availability in the coming weeks.
The signal for builders: the lab you bet on for the next 18 months is no longer the obvious default. If your stack is OpenAI-only, this is the week to test Claude on your actual workload — not because it's hype, but because the unit economics, the talent flow, and the coding benchmarks all flipped at the same time.
Microsoft Build's real story: they built their own coding model to take Copilot back from Claude.
At Build (Jun 2–3), Microsoft unveiled MAI — homegrown models including a coding-specialized variant aimed at one target: reclaiming GitHub Copilot. The framing is direct because the data is. Claude Code overtook Copilot as the dominant developer AI tool earlier this year. MAI is the response.
What shipped:
→ MAI model family — coding, transcription, reasoning, speech, image. In-house. Bundled into Copilot.
→ Agent Framework for .NET and Python — production-ready, multi-agent orchestration
→ Multi-model Copilot platform — now includes Anthropic models. Microsoft hedging openly.
→ Azure AI Foundry updates — cost governance, token monitoring, policy enforcement at the platform layer
→ Windows AI SDK — unified ONNX Runtime + DirectML + Copilot Runtime in one NuGet package for local AI
The strategic read: Microsoft now ships its own models, OpenAI's models, and Anthropic's models inside Copilot. Foundry routes across all of them. This is not loyalty — this is optionality. Microsoft is positioning to be vendor-neutral infrastructure while the labs fight.
The builder move: if you build on Azure, Foundry's cost governance just became the most important page in the docs — track which model you're actually paying for per call. If you build on Windows desktop, the unified AI SDK collapses three integrations into one. If you compete with Copilot, MAI's coding model is your new ceiling to beat.
OpenAI is shipping defense — Codex Computer Use on Windows + a GPT-5.5 quality patch.
While the valuation crown changed hands, OpenAI shipped concrete builder updates the same week.
Codex — May 30:
→ Computer Use on Windows — Codex can now see, click, and type inside Windows apps. Eligible Mac users get locked computer use.
→ Remote continuation — start a task on desktop, continue from mobile or Mac
→ Codex Profiles — usage and token activity per profile
→ Appshots — attach an app window to a Codex thread with a hotkey (Mac); Codex reads the screenshot + visible text without setup prompts
→ Goal mode GA — define an outcome + success criteria across CLI, IDE, and app
GPT-5.5 Instant update:
→ Clearer, more natural responses (less bullet-heavy)
→ Writing blocks and code blocks supported directly in chat
→ Pacing improvements on practical help tasks
The read: OpenAI is methodically closing the Codex experience gap with Cursor + Claude Code on the surfaces it controls. Computer Use on Windows is the unlock for enterprise — the OS your finance team runs is now agent-addressable.
Worth testing this weekend: if you've ignored Codex since the early ChatGPT-in-VS-Code days, the Goal mode + Appshots combination on Mac is the closest OpenAI has come to a real agentic dev loop. Run one refactor against it, measure against your Claude Code baseline.
Forward look: WWDC tomorrow, Groq pivots to inference, and the EU AI Act deadline is 8 weeks out.
→ WWDC keynote — Monday, Jun 8, 10 am PT
Apple's main 2026 AI moment. Expected: Siri 2.0 (chatbot-style, far more capable than current Siri) and a multi-model AI selector letting users pick third-party assistants (Claude, ChatGPT, Gemini) for Writing Tools and Image Playground. Apple Intelligence is pivoting from a closed system to a platform. If your app exposes App Intents, this is a distribution unlock.
→ Groq raising $650M — pivoting to inference
Reported May 29. Groq is moving from chip hardware to AI inference services — refining how models respond to prompts. Comes after Nvidia's $20B "not-acqui-hire" of an undisclosed AI chip company. Cheap inference is the next battleground. If you run agents at scale, watch the inference layer the way you watched model prices in 2024.
→ EU AI Act fully applicable Aug 2, 2026 — 8 weeks out
For Annex III high-risk systems (employment, credit, education, law enforcement): conformity assessments completed, technical documentation finalized, CE marking affixed, EU database registration done. Penalties: up to €35M or 7% of global annual turnover, whichever is higher. SME/SMC simplified documentation is now extended.
→ One more to test: Anthropic's 28 new enterprise security integrations — CrowdStrike, Palo Alto, Okta, Cloudflare, Wiz, Zscaler, Fortinet, more. Claude now governable inside corporate IT. If you've been blocked from selling Claude into regulated buyers, the integration list just got short.
Try These Before WWDC's Siri 2.0 Lands Monday
The vendor stack moved this week. Catch up before Monday's next move. Pick two:
☐ Run one real task against Claude Opus 4.8 — Pick something your current model gets wrong 1 in 3 times. Run it against Opus 4.8. Measure. 20 min.
☐ Spin up a Dynamic Workflows test — If you have a repo-scale migration sitting in the backlog, prototype it with Claude Code + Opus 4.8's 1,000-subagent mode. 45 min.
☐ Test OpenAI Codex Computer Use — On Windows if eligible, or Goal mode + Appshots on Mac. Run one refactor against your Claude Code baseline. 30 min.
☐ Audit your EU AI Act exposure — If you ship to EU users in employment, credit, education, or law enforcement contexts: list every Annex III touchpoint. The Aug 2 deadline is 8 weeks out. 20 min.
Don't bookmark. Don't "save for later." Pick two. Start today. →
AI Agents Are Reading Your Docs. Are You Ready?
Last month, 48% of visitors to documentation sites across Mintlify were AI agents, not humans.
Claude Code, Cursor, and other coding agents are becoming the actual customers reading your docs. And they read everything.
This changes what good documentation means. Humans skim and forgive gaps. Agents methodically check every endpoint, read every guide, and compare you against alternatives with zero fatigue.
Your docs aren't just helping users anymore. They're your product's first interview with the machines deciding whether to recommend you.
That means: clear schema markup so agents can parse your content, real benchmarks instead of marketing fluff, open endpoints agents can actually test, and honest comparisons that emphasize strengths without hype.
Mintlify powers documentation for over 20,000 companies, reaching 100M+ people every year. We just raised a $45M Series B led by @a16z and @SalesforceVC to build the knowledge layer for the agent era.
Forward this to one builder who's still defaulting to OpenAI because "Anthropic is the smaller one."
Reply with one word: Anthropic, OpenAI, and Microsoft — which stack will you be defaulting to by the end of Q3?