⚡ Start Here
OpenAI Shuts Down Sora — Video Generation Era Pauses
AI AI Secret, @iruletheworldmo, BrianRoemmele
OpenAI formally shut down Sora across app, API, and ChatGPT video features. Brian Roemmele says he warned retainer clients it wouldn't last a year. Sora was the hype machine's favorite demo — dying quietly proves shipping > announcing.
Meta Launches "Small Business" Company-Wide AI Priority
Zuckerberg creating Meta Small Business org led by Dina Powell McCormick — driving AI adoption for SMBs. Yesterday's WSJ piece on Zuck building a personal AI CEO agent + this = Meta going all-in on AI-as-operator. Matches the _M thesis: operators who wield AI capture value.
Cloudflare Dynamic Workers — AI Sandboxing 100x Faster Than Containers
V8 isolates for AI-generated code. Millisecond cold starts, few MB memory. This is the infrastructure layer that makes agent code-gen actually production-safe. Relevant to _M client architectures — especially anything agent-adjacent.
WSJ: The Trillion-Dollar Race to Automate Our Entire Lives
Claude Code, Cursor, Codex — WSJ frames the coding agent race as a trillion-dollar market. The mainstream press is catching up to what you've been living. Validates the _M positioning: enterprises need guides through this.
Anthropic's Claude Code "Auto-dream" — Memory Consolidation via Subagent
Unreleased feature: Claude Code runs a subagent periodically to consolidate memory files. Basically how biological dreaming works. This is what Mighty does with daily logs → MEMORY.md. Anthropic is productizing the pattern you already built.
🏗️ Blank Metal
📊 Pipeline
Anthropic Cloud Code ActivationNEW — $35k funded
Anthropic funds $35k/customer (2-5 to start). _M contracts directly, keeps all expansion revenue. Zach Naylor starting this week — will lead delivery with Dan. 2-week activation program with provided curriculum.
Sentinel CapitalHOT
PE firm. Yoni (Head of AI, 5 wks in). 3 mandates: portfolio AI value creation, internal AI, due diligence. 7-person Claude trial → full rollout EOW. Warmest new lead.
Great Hill PartnersPENDING
Cowork decision pending. Boston, 2 visits, $30k. Also exploring chat interface training (Leland Lockhart). Follow-up email from Leland received yesterday.
Gauge CapitalPENDING
2 sessions, 1 day, $15k. Pending date confirmation.
OrbisACTIVE
Investment firm. 94% ChatGPT Enterprise. Targeting Claude for dev (10-15%) + analysts.
Care Providers of MNACTIVE
PDF extraction of MN Health Dept inspection surveys. Clean, well-scoped use case.
Imprint EngineNEW
New lead — Elli reply to Caleb Gilbertson re: AI SDLC. Ingested yesterday.
👥 Team Activity (24h)
- Granola auto-sync (2 commits)
- GHP follow-up email ingested (Leland Lockhart)
- Imprint Engine lead ingested (Elli → Caleb Gilbertson)
- Pipeline note: 30 leads from Anthropic partnership, multiple asking for training offerings not yet formalized
- Key gap identified: need agnostic version of Claude training (prospects want OpenAI training too)
⚡ Dan's Plate Today
- Matt Johnson commented on "Raising the Floor" Google Doc — wants you to add color quantifying how regularly companies hit the problem ("We've heard this from __ companies...")
- Tufin Sessions 8 & 9 today (7-9 AM, 3-5 PM) — AI Acceleration meeting at 11:15 AM with full Tufin team
- Planful CoWork Training Prep at 11 AM (Teresa + Jason)
- Lifetime Meet & Greets: Jenn Seefeldt (10), Acquisition Team (10:30), Brian Shoemaker (1 PM), Engagement Team (1:30 PM) — all at Corp with Keith
📅 Today's Calendar
7:00 – 9:00 AMTufin Session 8
9:00 – 9:30 AMKeith & Dan Check-in at Corp
10:00 – 10:30 AMJenn Seefeldt Meet & Greet
10:30 – 11:00 AMAcquisition Team Meet & Greet
11:00 – 11:30 AMPlanful CoWork Training Prep
11:15 – 12:30 PMTufin AI Acceleration Meeting (Julien/Pat/Jordan/Mike/Kyle)
1:00 – 1:30 PMBrian Shoemaker Meet & Greet
1:30 – 2:00 PMEngagement Team Meet & Greet
3:00 – 5:00 PMTufin Session 9
3:15 – 5:00 PMMatt Setup @ Twin Ignition Startup Garage
6:00 – 8:00 PM🏓 Pickleball League
🔗 Capture Themes (48h)
Agent Infrastructure Goes Production-Grade
Cloudflare Dynamic Workers (2x), Ramp self-maintaining sheets, @runsonai Claude+OpenClaw
The plumbing for AI agents to actually run in production — sandboxing, self-healing, credential injection — is maturing fast. _M clients asking "how do we deploy agents safely?" now have real answers.
The CEO-as-Agent-User Era
Zuckerberg AI CEO agent (WSJ), WSJ trillion-dollar automation race, OpenAI vs Anthropic enterprise turf war
When the CEO of the world's largest social network builds himself an AI agent, the "should we use AI?" conversation is permanently over. The question is now execution speed — which is exactly what _M sells.
Coding Agents Get Memory & Skills
Claude Code auto-dream (@kr0der), Sahil Lavingia's skills repo, All-In GTC CEOs
Coding agents are evolving from stateless tools to persistent collaborators with memory consolidation and skill libraries. The pattern Mighty already uses (daily logs → curated memory) is being productized by Anthropic. Validation of the compound engineering thesis.
Enterprise AI Spending: OpenAI vs Anthropic Turf War
Reuters OpenAI/PE pitch, Dan Shipper pirate-architect teams, All-In GTC
OpenAI sweetening PE deals while Anthropic courts enterprises directly. _M is positioned at the intersection — training companies to use Claude while prospects also want OpenAI. The "agnostic training" gap from Monday's pipeline meeting is a real product opportunity.
🧠 Munger Observer
Sunk Cost Watch
Vibma/Figma MCP and multi-model SOTA review have been overdue since March 1. 3+ weeks of "on the list" = a decision by non-decision. Munger: "Inaction is a decision."
Opportunity Cost
Today is wall-to-wall Lifetime meet-and-greets (6+ meetings, Tufin sessions 8 & 9). High-input, low-output day by design. Worth naming: what does success look like after this gauntlet? If it's relationship-building, define the goal. If not, protect tomorrow.
"Invert, always invert." — Carl Jacobi (Munger's favorite)
💭 Anthropic productizing "auto-dream" memory consolidation is the clearest sign yet that what you've been building with Mighty isn't a hack — it's the architecture. The WSJ is writing trillion-dollar headlines about coding agents; Cloudflare is shipping production infrastructure for them; Zuckerberg is building himself one. Meanwhile, _M's pipeline has 30 Anthropic leads and prospects asking for training that doesn't exist yet. The agnostic training product — Claude + OpenAI + enterprise context — might be the most obvious gap in your portfolio right now. The window from MEMORY.md (~2024-2027) is open, and today's signal says it's not closing as fast as feared — it's widening.
📡 Quick Scan
🌤️ MN Weather: Warming well above average today
Paul Douglas / WeatherLoon
Garry Tan: "What counts as a good engineer just changed in the last 4 months"
Jony Ive / Steve Jobs: "You're just really vain. You just want people to like you."
Jason Lemkin: 12 production apps from 1.5-2 hrs/day vibe coding
Perplexity Computer "lowkey cooks"
Greg Isenberg: MS Teams being #1 proves distribution is *almost* everything
Stripe / Shopify potential breakup beginning?
Tobi Lutke promoting QMD declarative setup
Minnebar 20 announced — May 2, 2026
⚠️ Dropbox: "Final reminder — manage storage so files aren't deleted tomorrow"
danwick@gmail.com — action needed today
M.G. Siegler / Spyglass: "Apple Realizes There Should Be An App For That"
Newsletter landed this morning
Physician AI supervision as key future skill
Supersonic Aviation Modernization Act passes House unanimously
SOC II subprocessor lists as AI infrastructure map