💪 Mighty Brief

Monday, March 30, 2026 · 6:17 AM CT

🎯Start Here

OpenAI kills Sora, redirects compute to Codex
HOT WSJ · TechCrunch · Techmeme top story
WSJ details the rise and fall of Sora — isolated from research team, burned compute, shut down alongside the "adult mode" feature. GPUs now grinding on Codex. The real story: OpenAI is in full panic mode racing Anthropic on enterprise/coding. Freed GPU = Codex scaling. WSJ · TechCrunch
Lenny's Newsletter: "How OpenClaw Changed My Life" — Claire Vo
SIGNAL Lenny Rachitsky
Claire Vo runs 9 OpenClaw agents on Mac Minis. Journey from skeptic (it deleted her calendar day one) to building her entire business/podcast/family life on it. Multiple specialized agents > one generalist. This is the OpenClaw mainstream moment — Lenny has massive PM audience reach. Read
Garry Tan: 37K LOC/day across 5 projects, open-sourced his workflow
SIGNAL YC President, X
72 days into agentic coding — pacing at 90x his 2013 output. GStack browser now has multi-tab + context extraction back to Claude Code. Open-sourced everything. "I'm literally building every idea I had today instead of putting it in Apple Notes." Thread · Open source
Jason Lemkin: "We went from 8 humans in sales to 1.25 — agents chase the small deals"
HOT SaaStr
Agents don't mind chasing small deals. They never get tired. Same pipeline throughput with 85% fewer humans. This is the M thesis playing out in real enterprise sales data. Link
Artemis II 'GO' — four astronauts to the Moon, Wednesday launch
HOT NASA
Zero technical issues. First crewed lunar mission since 1972. Wednesday launch. SpaceX hype is also boosting space stocks across the board (The Information). Link

🏗️Blank Metal

📅 Today's Calendar (Heavy)

8:00 AMHOLD — Intro Call | Bit Farms (pending confirmation)
8:30 AMDeal Flow Check-In w/ Elli
10:00 AMPlanful Coffee & Claude Training (4 hrs, ~30 attendees)
10:30 AMWeekly _M Pipeline & Staffing (overlaps Planful)
2:00 PMTufin Working Session #12 — Executive Dashboarding
2:30 PMDan ↔ Mike Sanders — Tufin Renewals Use Cases
3:00 PMTufin Session #11 — Customer Intelligence & O2 Decks
3:30 PMHOLD — Intro Call | Bit Farms
5:00 PM🏓 Pickleball 4.0 League

⚡ Dan's Plate

  • 📧 Teresa shared HR&A SOW doc — "HRA Claude-Cowork-Training-SOW" needs your review. Proposal was due this morning per MEMORY.md. ACTION
  • 🎓 Planful training (10a-2p) — First Claude training for their team. Coffee & Claude format. ~30 attendees. Foundation + Application + Open Q&A.
  • 🔧 Tufin triple-header (2-4p) — Executive Dashboarding, Mike Sanders renewals walkthrough, Customer Intelligence. Back-to-back-to-back.
  • 📞 RudderStack/Omnia renewal sync (2:30p) — conflicts with Tufin; may need to delegate or reschedule.

📊 Pipeline Snapshot

HR&A Advisors🔥 SOW due TODAY
Tufin✅ Active — 3 sessions today
Planful✅ Training today
Sentinel Capital🟠 Claude rollout EOW
Tungsten Automation🟠 Awaiting Celia Moore
Great Hill Partners🟠 Cowork decision pending
Orbis🟠 94% ChatGPT, targeting Claude for dev/analysts
Bit Farms🟠 Intro hold today
Care Providers MN✅ PDF extraction scope
Edmunds GovTech🟠 AI expansion convo

🔗Capture Themes (48h)

Agent Infrastructure Is the New Platform War
Sources: @karrisaarinen (agent context), @auchenberg/Brandur (APIs as agent portals), @systematicls (long-running agent harness), @_chenglou (Pretext — DOM-free rendering)
Why it matters: Four practitioners converging on the same insight: the agent's context layer is the product, not the model. Directly validates OpenClaw's architecture and _M's "System of Intelligence" thesis.
Compounding Through Micro-Experimentation
Sources: David Senra × Tony Xu (DoorDash), Alfred Lin (experimentation compounding), Tim Ferriss × Jim Collins (return on luck, punch cards)
Why it matters: Three separate sources landed on the same mechanism: small bets that compound > big swings. DoorDash ran hundreds of thousands of experiments/week. Jim Collins' "return on luck" is the same idea wearing a different jacket. Maps directly to T3 (compounding engineering as moat).
The Anthropic Breakaway
Sources: All-In E266, Garry Tan × Boris Cherny (Claude Code origin story), OpenAI Sora shutdown
Why it matters: All-In quantified it: Anthropic added $6B ARR in February alone. OpenAI is killing products (Sora, adult mode) to redirect compute to Codex — defensive move against Claude Code dominance. Garry Tan attributes his own productivity transformation to Boris Cherny's Claude Code philosophy. _M's bet on Claude is looking increasingly right.
Venture Model Under Pressure
Sources: Gili Raanan/20VC (venture model broken, margins don't matter), All-In E266 (SaaS apocalypse, OpenAI offering PE 17.5% guaranteed returns)
Why it matters: The old growth metrics don't work in AI — Raanan says margins don't matter early stage, growth rate requirements have shifted. Wick.tech investment thesis needs to account for this: AI-native startups look different from SaaS vintage.

🧠Munger Observer

Opportunity Cost / Sunk Cost: Two tasks (Vibma, multi-model research SOTA) have been due since March 1 — 25 days overdue. Invert: if they haven't been done in 25 days, ask whether they should be done at all. Overdue todos that stay overdue are decisions pretending to be deferrals.


Second-Order: Meeting Compression. Tomorrow stacks 5 meetings before noon. First-order = get the meetings done. Second-order = every meeting is a commitment factory. Who's capturing what was promised before the next sprint of meetings lands?


"Invert, always invert." — Jacobi (Munger's favorite)

📰Newsletter Inbox

Every: "Everyone Gets a Sidekick"
every.to · Mar 29
The AI sidekick wave is mainstream enough for Every to cover it as inevitable. Read
The Information: SpaceX Hype Boosts Stocks in Its Crosshairs
theinformation.com · Mar 29
SpaceX's soaring valuation + potential IPO is lifting all space-adjacent stocks. Read
Lenny's Community Wisdom: AI velocity outpacing product strategy
lennysnewsletter.com · Mar 28
Community thread on when AI shipping speed outruns strategy. Relevant to _M clients wrestling with adoption velocity. Read

⚡Quick Scan

💭Personal Thought

Garry Tan is 72 days into agentic coding and pacing 90x. Claire Vo runs 9 OpenClaw agents for her business. Jason Lemkin documents 85% sales headcount reduction via agents. Three data points from three different domains, all landing in the same 48 hours. The "AI-native or die" thesis (T1) isn't a prediction anymore — it's a scoreboard. The question for _M isn't whether clients need this. It's whether you can staff the demand. Today's calendar — 8 meetings, 3 Tufin sessions, a 4-hour Planful training — is evidence of a pipeline that's outrunning capacity. The Munger Observer's note about meeting compression is dead on: every one of those meetings is a commitment factory. Who's capturing what was promised?
Mighty Brief · Generated by 💪 at 6:17 AM CT · Archive