⚡TL;DR
- Yann LeCun launches AMI Labs — $1.03B seed at $3.5B pre-money. Europe's largest ever seed. Building world models as alternative to LLMs. Source
- China OpenClaw craze — Software stocks surging 20%, Shenzhen/Wuxi drafting subsidies up to $290K+. MIIT warning of "significant security risks." The tool running Mighty is moving Chinese stock markets. Bloomberg
- Spark Capital raising $3B — Anthropic's first VC backer, up 50% from prior funds. The Information
- OpenAI IPO skepticism — 11 public market investors surveyed, mood is "mixed." Could eclipse Saudi Aramco's $25B but 6+ months away. The Information
- "AI Was Supposed to Free My Time. It Consumed It." — Every piece on the intensity paradox. Direct hit on your "AI Intensifies Work" draft. Every
- Figma MCP server goes bidirectional — Full design → code → canvas → feedback roundtrip with GitHub Copilot. Dylan Field RT
- Stripe billing for LLM tokens — Trivial pass-through and markup of token costs, easing flat-tier to usage-based shift. Phin Barnes
- Ethan Mollick on AI alternative approaches — Half dozen well-funded companies from famous researchers betting LLMs hit a wall. More pathways than ever. Source
🔗Capture Themes (Last 48h)
Smart Money Preparing
@wolfejosh Lux Capital memo · The Information "Quiet Part Out Loud" · OpenAI IPO skepticism
Converging signals: VCs privately doubt AI valuations, urging founders to extend runway and stress-test assumptions. "Failure comes from a failure to imagine failure." Directly relevant to _M client conversations and Helios prep — smart money is hedging while still deploying.
Agents Eat Software
@levie "Building for Trillions of Agents" · Bloomberg China OpenClaw craze
Levie wrote the macro case for _M's entire consulting model: every client engagement is "make your systems agent-ready." Government subsidies + 20% stock moves + security warnings = agents crossing from developer toy to enterprise infrastructure. This is the wave _M rides.
Process > Atoms
Founders #414: SpaceX "Atoms Are Cheap, Process Is Pricey" · Levie API-first mandate
SpaceX thesis in physical form: materials = 2% of price, 98% is process waste. Same pattern in AI services — compute is cheap, integration/orchestration is expensive. Vertical integration + fast iteration + volume = moat. This IS the compounding engineering thesis.
🏗️Blank Metal
📊 Pipeline
| Deal | Status | Last Activity |
|---|
| Tufin | Active | Yesterday — Claude Code training, Elli coordinating licensing (23 users Wave 1). Anthropic can't call; EJ stepping in. |
| TritenIAG | Intro Today | Intro call 8:00 AM with Terence Easton, Matthew Harris, Darla Haseloff, Mickey Reeves |
| BBG | Lead | Mar 3 |
| Bond Sports | Lead | Mar 3 |
| Oshi Health | Lead | Feb 25 |
| SRI | Stale (13d) | Feb 25 |
| Talkiatry | Stale (13d) | Feb 25 |
| PointsNorth | Stale (15d) | Feb 22 |
| Dasvandh | Stale (8d) | Mar 2 |
👥 Team Activity (24h)
- Tufin session: renumber script for 59-slide deck, session debrief, updated learnings
- Granola auto-syncs (2x)
- Elli emailing Tufin re: Claude Code licensing — Anthropic team too stretched, EJ offering direct help
- Sheree Wong (Tufin): 23 team members for Wave 1 Claude Code training, evaluating Team vs Enterprise license
⚡ Dan's Plate
- 8:00 AM — Triten intro call (Zoom). New pipeline opportunity — Munger warned yesterday to not under-prep for this.
- 2:00 PM — AI Sprint structure retro (full team: EJ, Elli, MJ, Zach, Teresa, Josh, Emily)
- 3:00 PM — Mark & Dan checkin
- 3:30 PM — Dan/Rob Connect (Teams, personal calendar)
- 4:00 PM — Training logistics (Zach, Teresa) — coverage while Dan's out
- Elli/Tufin email thread: Tufin wants Anthropic pricing guidance for 23 users, EJ handling but may need your input
🧠Munger Observer
Opportunity Cost Alert: Heavy Helios prep day — call prep, two Resonance captures, pricing options, room-specific attendee notes. Meanwhile, Triten intro call is at 8am tomorrow. Are you over-indexing on the deal you already have traction on vs. the new pipeline? Helios prep looks thorough enough; tomorrow morning should include 15 min on Triten context.
Second-Order Check: AI usage tracker script is good first-order thinking (know what you spend). Second-order question: what changes when you see the numbers? If the answer is "nothing," tracking is theater. Define the threshold that triggers action before you have the data, not after.
— "Invert, always invert."
📅Today's Calendar
8:00 AMIntro Call — Blank Metal ↔ Triten (Zoom)
2:00 PMAI Sprint Structure Retro (Google Meet, full team)
3:00 PMMark & Dan Checkin (Google Meet)
3:30 PMDan/Rob Connect (Teams)
4:00 PMTraining logistics while Dan's out (Zach, Teresa)
📧Newsletters & Email
The Information · Katie Roof & Anissa Gardizy · Mar 9
Anthropic's first VC backer raising 50% more than prior funds. Tests investor demand in early AI winners. Follows Khosla's $4B+ raise (OpenAI backer).
The Information · Anita Ramaswamy & Miles Kruppa · Mar 9
At least 6 months from going public. 11 public market investors surveyed — mood is mixed. Could eclipse Saudi Aramco's $25B IPO. Banks already doing outreach.
Every · Mar 9
The intensity paradox — AI tools create more work, not less. Direct overlap with Dan's "AI Intensifies Work — Good" draft in the writing pipeline.
Lenny's Newsletter / Claire Vo · Mar 9
Jamey Gannon's workflow: style references > text prompts, personalization codes, editorial references as shortcuts. Key insight: "avoid prompting at all costs."
dan.wick@blankmetal.ai
Elli → Tufin: Claude Code Training Thread
Anthropic can't do a call (team too stretched). Elli offered EJ for pricing/plan guidance. Sheree confirms 23 users for Wave 1, evaluating Team vs Enterprise. Shay wants to understand plan differences.
View in Gmail
🤖Twitter — AI
@ylecun via @dedene · 3:26 AM CT
New EU AI lab "Advanced Machine Intelligence" unveiled. One of the largest seeds ever, largest for a European company. Working on world models, betting LLMs hit a wall.
@emollick · 2:36 AM CT
"More pathways than ever for keeping AI development moving forward." Multiple companies from famous researchers building non-LLM approaches.
@ryancarson · 11:15 PM CT
Running Symphony in ralph mode with concurrency=10. Adaptation of the Symphony framework for parallel coding agent orchestration.
@tszzl · 2:16 AM CT
"There are military secrets much worse guarded than whatever SSI is up to."
@fidjissimo RT @rohanvarma · 2:33 AM CT
Leverages frontier models for code review without $25/review pricing. Fidji Simo (Instacart CEO) signal-boosting.
💡Twitter — Leaders
@zoink RT @mariorod1 · 1:27 AM CT
Design → code → canvas → feedback → repeat. Full roundtrip workflow with GitHub Copilot.
@hnshah RT @jackbutcher · 12:52 AM CT
"You are not what you can make. You are what you can see."
@msuster RT @AGHamilton29 · 11:20 PM CT
"I really don't think people understand what a significant shift this is from the past 3 decades."
🛠️Twitter — Tech/Dev
Techmeme via Bloomberg · 2:50 AM CT
AI-powered factories requiring essentially no workers set to upend China's labor market already stressed by tariffs.
Techmeme via Wired · 1:01 AM CT
Making it easier/cheaper to start software companies + agentic investors analyzing pitch decks and teams.
@phineasb · 11:42 PM CT
Trivial pass-through and markup of LLM token costs. Easing shift from flat-tier to usage-based pricing across the application layer.
@clairevo · 10:48 PM CT
"Do you know how much I'd ship if I didn't care about code quality? Cursor AI bugbot just out here playing D."
@steipete RT @cnakazawa · 12:35 AM CT
Coding agents are going to consume dramatically fewer tokens when running Vitest soon.
@steipete RT @poezhao0605 · 12:31 AM CT
Zhipu launched AutoClaw (stock jumped 13%). List: Moonshot KimiClaw, MiniMax MaxClaw, Alibaba... Chinese companies racing to build agent platforms.
@steipete RT @NVIDIAAIDev · 12:11 AM CT
NVIDIA recognizing OpenClaw as top user of Nemotron 3 Nano 30B on OpenRouter.
🔧Twitter — Tools
@AravSrinivas · 12:52 AM CT
Tool built with Perplexity Computer to port Spotify playlists to YouTube Music by just pasting a URL. "Cross-service migrations going to be seamless."
@steipete RT @swyx · 12:17 AM CT
"If you can build a category leader open source project in AI engineering right now the market acquihire rate is ~$10-$100M per person."
🏙️Twin Cities
@ChrisRstrib (Star Tribune) · 5:59 PM CT
Dangerous Man Brewing returning to its old neighborhood with a taproom in HeadFlyer's popular NE Minneapolis space.
Finance & Commerce · 7:00 PM CT
Enclave Companies purchased the former Macy's Furniture site in Edina for $18M. 523-unit housing and mixed-use redevelopment planned.
Finance & Commerce · 5:00 PM CT
Rochester EDA approves $139M CityWalk Apartments near Mayo Clinic — city's largest multifamily project.
📰Headlines
🧩Trending Skills
- OpenClaw on NVIDIA Nemotron — Now the #1 user of Nemotron 3 Nano 30B on OpenRouter. Relevant to local inference strategy.
- Figma MCP bidirectional — Design ↔ code roundtrip. Relevant for any client-facing prototype workflow at _M.
- Vitest token optimization — 80-90% fewer tokens for coding agents. Direct impact on Codex/Claude Code costs.
- skills.sh top movers: nano-banana (image gen), agent-browser, twitter-automation, remotion-render, Microsoft Azure AI skills surging