In partnership with
Today on FindMeAI:
The plumbing of the AI economy got rebuilt this week — distribution, transactions, vertical agents, and compute, all in seven days.
→ OpenAI walks out of Azure — Microsoft's exclusivity dies; GPT-5.5 lands on AWS Bedrock the next morning.
→ Agents got a wallet — Cloudflare and Stripe shipped a protocol that lets agents buy domains, spin up servers, and pay for them.
→ Anthropic shipped 10 production agents — pitchbooks, KYC screening, month-end close — plus Claude inside Excel.
→ Anthropic rented all of SpaceX's Colossus 1 — 220,000 GPUs, and your Claude Code limits just doubled.
Microsoft and OpenAI tore up their original partnership and signed a new one. The headline change: Microsoft's exclusive right to host and resell OpenAI's models is dead.
→ AGI clause: gone. Replaced with a fixed-date IP license running through 2032.
→ Revenue share: Microsoft no longer pays OpenAI a cut on Azure usage.
→ Cloud lock-in: Microsoft's license is now non-exclusive.
The next morning, GPT-5.5 went live on AWS Bedrock as part of a separate $38B OpenAI–Amazon agreement. Google Cloud distribution opens next.
Then on May 5, OpenAI made GPT-5.5 Instant the default model for every ChatGPT user — free, Plus, and Pro. Internal evals show 52.5% fewer hallucinated claims on high-stakes prompts in medicine, law, and finance, and 37.3% fewer factual errors on prompts users had previously flagged.
What changed | Before | Now |
|---|---|---|
OpenAI cloud distribution | Azure-exclusive | Azure + AWS Bedrock + GCP |
ChatGPT default model | GPT-5.3 Instant | GPT-5.5 Instant |
Microsoft revenue share | Yes | No |
For builders: pick your cloud. The lock-in is over, and the strongest model in production just got cheaper to route around.
Cloudflare and Stripe shipped a joint protocol that lets AI agents create cloud accounts, register domains, deploy apps, and pay for it all — with no human in the loop beyond the initial Terms of Service.
The protocol has three layers:
→ Discovery — the agent queries a catalog of available services
→ Authorization — the platform attests user identity, issues credentials to the agent
→ Payment — Stripe issues a billing token; the agent can start subscriptions and run usage-based charges
Built-in guardrails worth knowing:
→ Raw card details are never exposed to the agent
→ Default Stripe cap: $100/month per provider (configurable)
→ All transactions traceable back to a verified user identity
This is bigger than Cloudflare. Coinbase, Cloudflare, and Stripe are spinning up a nonprofit foundation to govern x402 — an open protocol for machine-to-machine payments. It's competing with at least one rival standard to become the default rail for agent commerce.
For builders: the missing primitive — agents that can transact — just shipped. The next 12 months of agent products will be defined by who builds on this rail first.
Anthropic Just Shipped 10 Production Agents for Wall Street
On May 5, Anthropic released ten ready-to-deploy agent templates for financial services. Not demos. Not blog posts. Working blueprints that ship as plugins inside Claude Cowork, Claude Code, and as cookbooks for Claude Managed Agents.
Research & Coverage | Finance & Operations |
|---|---|
Pitch Builder | Valuation Reviewer |
Meeting Preparer | GL Reconciler |
Earnings Reviewer | Month-End Closer |
Model Builder | Statement Auditor |
Market Researcher | KYC Screener |
The KYC Screener ships with a kyc-rules skill that encodes how Claude applies a firm's KYC/AML policy to a parsed onboarding record — production logic, out of the box.
The bigger play: Claude is now native inside Microsoft 365. Add-ins for Excel, PowerPoint, and Word are live; Outlook is coming. The integrations share context across apps, so a model built in Excel flows straight into a PowerPoint deck without re-entering data. A Moody's MCP app shipped alongside.
For builders: this is the vertical agent playbook. Pick a vertical, ship 10 templates, embed inside the tools the customer already uses. Anthropic just published the recipe — copy it.YOUR MAY PLAYBOOK
5 Things to Carry Into May
Ignore the noise. These five compounds:
→ Watch your token bill weekly. NVIDIA's own VP said compute now exceeds salary cost on his team. Uber burned its 2026 AI budget by H1. If you're not measuring per-feature token spend, you're flying blind.
→ Build vertical, not horizontal. Avoca hit $1B valuation building agentic AI for plumbing and HVAC companies. The unsexy verticals are where real revenue lives. Pick one industry, go deep.
→ Audit your agent fleet. Count every agent in production. If you don't know the number off the top of your head, you're already in the 94%.
→ Pick a side on military AI. Anthropic refused the Pentagon. Google, OpenAI, and xAI signed. Your tools belong to one side or the other — know which.
→ Default to the best model, not the cheapest. Opus 4.7 is +13% on real engineering tasks. The cost of choosing wrong is now larger than the cost of running both.
Anthropic's New Compute Provider: SpaceX
Anthropic signed an agreement to take all of the AI compute capacity at SpaceX's Colossus 1 data center. The numbers:
→ 300+ megawatts of new capacity online within the month
→ 220,000+ NVIDIA GPUs — H100, H200, and GB200 accelerators
→ Capacity arrives ahead of Anthropic's reported June IPO
What it means for paying users — effective immediately:
→ Claude Code 5-hour limits doubled for Pro, Max, Team, and Enterprise
→ Peak-hours throttling removed for Pro and Max
→ Claude Opus API rate limits raised
Bonus signal: Anthropic and SpaceX are also exploring orbital AI compute — multi-gigawatt capacity in space.
Other tools that shipped this week:
→ Microsoft Agent 365 — control plane for enterprise agents. $15/user/month. Launched May 1.
→ Mistral Workflows — orchestration engine for moving AI from prototype to production with built-in observability.
→ xAI Grok Imagine 1.0 — image-to-video; currently #1 on DesignArena, beating Runway Gen-4.5, Sora 2 Pro, and Veo 3.1.
→ Snapchat AI Sponsored Snaps — brand AI agents inside the chat interface, replacing static ads.
The pattern across every story: the bottleneck this week wasn't models. It was access — to clouds, to wallets, to enterprise tools, to compute. That bottleneck just broke.
Try These Before Monday
Reading about new rails is one thing. Wiring something to them is the only way the muscle memory builds.
☐ Switch your strongest workflow to GPT-5.5 on Bedrock — point one production agent at the new endpoint and compare it to your Azure deployment. 45 min.
☐ Spin up a Cloudflare agent that buys its own domain — follow the Cloudflare-Stripe quickstart, set the $100 cap, ship a one-page site end-to-end. 1 hour.
☐ Install Claude's Excel add-in and rebuild one recurring report — pick the report you hate most. Let Claude take the first pass. 30 min.
☐ Re-run the heaviest task you put off this month in Claude Code — your limits just doubled. The thing that hit a wall last week probably won't now. Open-ended.
Don't bookmark. Don't "save for later." Pick two. Start today. →
Investors see ANOTHER return from Masterworks (!!!!)
That’s 6 sales in 7 months. 29 all time. And the performance?
16.5%, 17.6%, and 17.8%, net annualized returns on sold works held longer than one year (See all 29 at Masterworks.com)
It’s not from stocks, private equity, or real estate… it’s from contemporary and post war art. Crazy, right?
With Masterworks, you don’t need to be a BILLIONAIRE to invest in multi-million dollar art anymore.
Historically, the segment overall has had attractive appreciation and low correlation to stocks.*
Masterworks targets works featuring legends like Banksy, Basquiat, and Picasso, identifying what they believe to have significant long-term appreciation potential, not just at the artist level but at the level of individual artworks.
As one of the largest players in the art market, with $1.3 billion invested over 500 artworks, they pass critical advantages through to their 70,000+ members to add art to their portfolios strategically.
Looking to diversify your investments in 2026?
*According to Masterworks data. Investing involves risk. Past performance is not indicative of future returns. See important Reg A disclosures at masterworks.com/cd.
Forward this to the engineer on your team, who is still routing everything through Azure. They need to know.
Reply with: which of the four layers — distribution, transactions, agents, or compute — is going to bite your roadmap first? I read every reply.