A pretty bold comment in Nature written by linguists, computer scientists and philosophers declaring that AGI has been achieved. "By reasonable standards, including Turing’s own, we have artificial systems that are generally intelligent. The long-standing problem of creating AGI has been solved."

After executing Alex Pretti on Saturday, ICE has shifted to primarily targeting bus stops and elementary schools in my community this week. I cannot overstate how much of the ICE activity in my community is just targeting places where little kids are.

Auschwitz was at the end of a process. We must remember that it did not start from gas chambers. This hatred gradually developed: from ideas, words, stereotypes & prejudice through legal exclusion, dehumanization & escalating violence... to systematic and industrial murder. Auschwitz took time.

🧵 The stickiest AI products don’t feel magical. They feel inevitable. They don’t just show off what the model can do. They erase something frustrating from your workflow—so smoothly, you can’t imagine going back.👇

1/ 🔗 Founders are sleeping on one of the most powerful uses of AI right now: Prompt chains. 🔗 Not for writing emails or fixing grammar. But for thinking. Strategically. Here’s how I used one to generate startup ideas, GTM plans, SEO strategy—and more—in under an hour 🧵

Name a founder more people should be learning from—someone whose name doesn’t get mentioned enough. I’ve been lucky to back and learn from some remarkable founders. The ones I keep thinking about aren’t always the loudest or most followed. (1/3)

If AI could take one thing off your plate today -- what would you hand over without hesitation? I’ve been asking this in 1:1s with founders. And what’s surprising isn’t that they’re automating more -- it’s what they’re automating first. (1/3)

Everyone’s watching AI demos. But the real AI startups? They’re compounding quietly. I went deep on the Lean AI Leaderboard — dozens of startups with breakout traction. Here are 10 patterns from the teams actually pulling ahead 🧵

None of these existed in 2020. Anthropic: $61.5B Cohere: $5.5B ElevenLabs: $3.15B Harvey: $5B Moonshot AI: $3B Together AI: $3.3B Figure AI: $39.5B Crusoe: $2.8B Character.AI: $10B Perplexity AI: $18B Adept AI: $20B Mistral AI: $6.2B Combined valuation: $178 billion. #MondayMotivation

Everyone jokes about the 10X engineer. But the best ones? They absolutely exist—and they’re evolving. 🧵 The New Engineering Stack in the age of AI:

VC is broken. AI killed the old playbook. Here’s what’s next 🧵

Maybe we should all post inventions and innovations for a week, “made possible by NIH” or other grants. Science is historically not good enough at self promotion and this is a storytelling war.

@parulia.bsky.social distills to the core

In a world where software is a commodity and GPT-4 is table stakes… What actually gives startups a moat? Let’s talk about two of the most powerful—but often misunderstood—defenses: Data moats vs. Brand moats. 🧵

Embedded AI’s quietly winning—like spellcheck sneaking into Word and ruining standalone’s day. Thought UX didn’t matter? It’s everything now. Standalone apps are out there fighting gravity.

AI agents are here—CB Insights shares the wild stat that 60% of AI startups in 2025 are building self-running systems. Why now: ✅ Agents handle tasks solo 🤖 Scale without humans 🏎 Outpace LLMs fast

VC myths are killing your raise—here’s the truth: 📈 Raw traction beats polish (Notion’s buzz) ❤️ User love outpitches you (Figma’s advocates) 🔎 10 meetings > 100 (focus wins)

Big LLMs are overhyped—small, adaptive AIs are the real play: 🧬 Evolves mid-task (xAI’s tech) 🎥 Redefines creative (Runway’s films) 🎉 Quietly wins (no fanfare)

AI’s not a tool—it’s a boss. Jasper’s $100M ARR in 2024 proves it: ✍️ Writes ads solo 📈 Funds its growth 🛞 Humans just steer

RL’s stealing the AI show—ArXiv says it cuts costs 40% vs. LLMs. Why it’s huge: Adapts mid-task (Tesla’s Optimus) Outsmarts static models Wins where Big Tech lags

VCs are pickier than ever—PitchBook says Series A raises now take 5-7 months. Stop the grind: 👓 Focus beats volume 🪝 Lean pitches hook fast 👯 Network trumps cold emails

VCs are done blindly chasing growth—First Round says 70% now prioritize capital efficiency. Here’s why it's huge: 💰 Revenue > hype 🥇 Lean teams win 🤑 Profit signals trust The hidden truth? Relying on more funding adds risk. Capital independence is the quiet edge. Not sexy, but it works.

Big LLMs are over—ArXiv says micro-models like Mistral’s Mixtral 8x7B beat them on niche tasks with 10x less compute. Why it matters:

Him: I've been reading a lot of post-apocalyptic stories lately. Me: Oh, you mean the newspaper?

AI-native companies create compounding moats: Data flywheels, model improvements, and automation loops. SaaS? Mostly UI with APIs. Big difference.

This is spending I would have been happy for my tax dollars to support. Local farms + schools and food banks: seemed win/win www.politico.com/news/2025/03...

New world order: Software = commodity? New platforms > old ones GTM + data = moats Early enterprise deals = consulting contracts Tiny teams -> profitable fast No one wants to IPO bc public markets are fickle, risky & onerous. Founders know VC rules & can fund each other. Growth VC may be redundant.

Venture dynamics are tricky for founders to navigate, until they’ve been through multiple rounds of funding and sometimes an acquisition. 👍🏽 10 rules of thumb for savvy founders: 🧵

My mother taught me everything about resilience, execution, and capital efficiency—long before I ever worked in startups. She never thought of herself as an entrepreneur. But in every way, she was one. 🧵

If everyone who did #StandupForScience didn’t also call their congresspeople it will have been pretty toothless. Given non-participation rates they only care about voters if they are demonstrably engaged. 🧬🦠🇺🇸

Don’t try to do everything and build everything for every customer — As tempting as it may seem to hedge your bets, it doesn’t enable you to build your own unique differentiation over time. Yes, this means you have to choose (and be stronger for it).

All the lenses on qualifying sales leads apply to finding P/M fit: —> who’s the buyer —> what are their purchase considerations —> is this a top 3 problem for them —> what could block a sale —> is it 10X rather than 2X better —> does it overcome inertia

Startups don’t fail because of bad ideas. They fail because: ⏱ they execute the right idea at the wrong time 🤺 the cofounders fight and split up 🚚 the founders never get a handle on distribution Testing market pull as early as you can fixes 2/3 so don't skimp on it.

AI won't replace managers. But it will force bad managers to improve—fast. 📨 A manager who just summarizes reports & sends emails? AI can do that. 👏🏽 A manager who hires, mentors, and builds teams? AI enhances them. People skills are the last *real* moat in leadership.

Some founders get caught up on ideas. But execution is what separates a billion-dollar startup from a good - even great - idea. 🔎 Google wasn’t the first search engine. 📘 Facebook wasn’t the first social network. 💲 Stripe wasn’t the first payments startup. They just out-executed everyone else.

Hackers love to hack systems, and sometimes the altruistic ones try to hack systems that will better the world. I’ve done this myself. But technical systems are harder to hack, than human ones. Healthcare, education, public policy, the problem of the commons —> not so easy to hack.

Remember that getting profitable gives you more options than not. @garrytan calls it getting into cockroach mode. crawl, walk, run, then race. (*hypergrowth is the race.)

VC hot take: unsexy markets are actually the hottest ones

Sim2Real + reinforcement learning + better synthetic data means robots will get *way* better without human-labeled training data 🤖

Move past SaaS metrics. In AI-native businesses, we ask: 🚀 Data flywheel speed? 🛠️ Cost per inference? 📉 Decreasing latency? 📊 Model efficiency gains? Growth isn't just ARR—it's intelligence accumulation.

The biggest AI companies in 2030 won’t be building models. They’ll be **controlling the data pipelines.**

Not enough people are talking about this insane advance that dropped last week. 🚀 AI just leveled up in synthetic biology. Scientists built the biggest AI model yet for designing DNA—moving from tweaking proteins to designing entire genomes. Here’s why this is so wild:

Your first 100 customers matter more than your first $1M in revenue. Why? Because they tell you: 🥰 If people actually *need* your product 💰 What they’re willing to pay 🎤 How to position your messaging Without strong early adopters, scaling is a guessing game.

People are out there building AI startups like it’s 2015 SaaS. It’s not the same game. The AI playbook is different: - There is no long-term moat in APIs. - The hardest part isn’t the model—it’s *distribution*. - AI products that feel ‘magical’ win. The rest get ignored.

If AI does 80% of the “hard” tasks in law, finance, and engineering… what actually makes someone a top performer in those fields? Curation Debugging Context Client management Something else?

Fundraising red flags: 5 signals that make VCs walk away 🚩 This week, I hosted a session with 5 founders prepping to fundraise. Their main question? What red flags make investors walk away? Here are 5 dealbreakers that can kill your raise before it starts: 🧵

If you’re not hiring the best people, you can be sure that your competitors are :-)

Knowing your own blind spots is truly a founder superpower. “I’m really good at product, but need to be supplemented on sales,” or vice versa. Hiring against those superpowers is a genius-level hack.

So many of my conversations with founders keep cycling back to the basics: ✅ simplify our language and messaging ✅ build trust and credibility with potential customers ✅ figure out the elements of a delightful experience ✅ reduce time to value ✅ run smart experiments ✅ do less but smarter