Alexandr Wang spent Scale AI’s first years walking into pitch meetings with great numbers and real revenue and getting rejected anyway.

Not one of the VCs across the table had ever trained a model. Data, in his words, was still “very unsexy.” A decade later, the same investors who passed on him write think pieces about how data is the biggest opportunity in AI.

Now Wang runs Meta’s Superintelligence Labs.

Before we get into his full conversation with Garry Tan at Y Combinator’s Startup School 2026, here is Scale AI’s funding journey:

• 2017, Series A: $4.5M raised at an undisclosed valuation, led by Accel.

• 2018, Series B: $18M raised at an undisclosed valuation, led by Index Ventures, with Accel and Y Combinator participating.

• 2019, Series C: $100M raised at a valuation of more than $1B, led by Founders Fund, with Accel, Coatue, Index Ventures, Spark Capital, and Thrive Capital.

• 2020, Series D: $155M raised at a valuation of more than $3.5B, led by Tiger Global.

• 2021, Series E: $325M raised at a $7.3B valuation, co-led by Dragoneer, Greenoaks Capital, and Tiger Global.

• 2024, Series F: $1B raised at a $13.8B valuation, led by Accel, with Amazon, Meta, Nvidia, and a long list of strategic and institutional investors.

• 2025, Meta deal: Meta invested $14.3B for a 49% stake, valuing Scale AI at more than $29B. Scale remained independent, while Wang joined Meta to lead its superintelligence push.

We went through his full conversation with Garry Tan and pulled out the parts worth stealing: the first-principles test that told him data was the wedge before anyone else saw it, the conviction playbook that explains why the smartest investors in the room were wrong for years, the exact agentic-loop mechanics that now beat a team of 100 engineers.

Bonus: at the end of the newsletter, you will also find Scale AI’s Series C pitch deck!

Inside you will find:

1. How Scale Really Started

2. Conviction Before Consensus

3. David vs Goliath Is Over: How AI Agents Turn Your Startup Into a Mecha-Goliath

4. How Wang Rebuilt a Frontier Lab in 9 Months

5. Vision Beats Intelligence: Why Ambition Is the Scarce Resource Now

6. The Agentic Loop That Beat 100 Engineers

7. What to Steal From Wang

8. Bonus: Scale AI Series C Pitch Deck

   

Launch your site faster than ever with Framer - now with Agents that make every part of your workflow effortless

Framer is the pro website builder for creators, teams, and businesses that care enough to get every detail right. Now with Agents that work alongside you across the full website workflow: designing new pages, managing your CMS, editing content, and adding SEO all without leaving the tool where the real site lives. Trusted by teams at companies like Miro and Perplexity.

Agents bring speed and scale; you bring taste, judgment, and control. Together you can:

• Build and iterate faster: generate sections, refine layout and styling, and ship updates in minutes.

• Scale content with confidence: update CMS collections, localize copy, and keep metadata consistent across pages.

• Stay production‑ready: reviewable changes, collaboration, and safeguards that help teams ship without handoffs.

Launch Your Site With Framer Today!

This issue + Scale AI Pitch Deck is available exclusively to paid subscribers!

Paid subscribers unlock the full library: 200+ decks, 10,000+ investors, AI Agent playbooks, and fresh high-signal content every week.

Not subscribed yet? Try it free for 7 days and unlock this playbook + Pitch Deck, plus our full premium library:

Get 7-Day Free Trial

We’re already 80,000+ founders and operators inside!

Let’s begin!

Read more

For 200 years, technology has automated cognition.

Spreadsheets replaced clerks. Software replaced accountants. AI is replacing analysts.

But the largest economic layer in the world has barely changed: manual labor. Warehouses, security, logistics, facilities, care work. Tens of trillions of dollars in wage spend every year, still dependent on physical human presence.

Robotics has tried to enter this space for decades. It failed because robots were:

• Too rigid

• Too unsafe

• Too expensive

• Too hard to deploy in human environments

1X is attempting something different. They are building general-purpose humanoid robots, safe enough to operate alongside humans, deployable in real commercial environments today, and now shipping into homes too. They’re building two different products:

1. EVE is 1X’s industrial platform, already deployed in enterprise settings like security and logistics.

2. NEO is 1X’s home robot: lighter, softer, and built for people’s living rooms rather than warehouses. It started shipping to its first Early Access homes in 2026, priced at $20,000 or $499 a month.

But here’s the deeper shift. If physical robots can be safely deployed at scale and if teleoperation remains part of the control stack, you don’t just automate labor. You create a labor network.

This is my vision of what this could mean:

Imagine something structurally similar to Fiverr, but instead of freelance design or code, it’s embodied work.

A worker in Manila, Nairobi, or Bogotá remotely supervises or intermittently controls a humanoid deployed in Dallas, London, or Tokyo.

Instead of exporting digital services, this unlocks the exporting and importing of physical labor.

Powerful models are only as useful as the company context they can access!

Make enterprise AI actually work with Glean, the context layer that connects every model, tool and workflow.

Glean connects your organization’s knowledge, people, permissions and activity to give AI the understanding it needs to deliver relevant, secure and work-ready results.

With one governed context and intelligence layer across tools like Claude, Cursor, Copilot, and custom agents, Glean helps your company:

• Give AI the right context: connect knowledge across 275+ enterprise applications while preserving permissions, metadata and access controls

• Get more value from every model: improve retrieval, reduce unnecessary tool calls and use approximately 30% fewer tokens than off-the-shelf MCP tools.

• Stay flexible as AI evolves: access 35+ models (including Claude, GPT and Gemini) without rebuilding your context layer or locking your company into one provider

Try Glean Now!

This is the industry this opens: Remote physical labor infrastructure.

And this is why this is genuinely one of the best robotics decks I’ve ever seen 🔥

1X isn’t pitching a flashy humanoid. They’re pitching a structural shift in how labor is produced and distributed.

Backed by:

• OpenAI Startup Fund (led their $23.5M Series A2)

• EQT Ventures (led their $100M Series B)

• ~$125M+ raised behind embodied AI

The deck portrays a very disciplined path from commercial wedge to labor platform.

This isn’t a robotics demo deck. It’s an industry-creation deck.

Now let’s break this beast down slide by slide 🔥

The pitch deck is available exclusively to paid subscribers!

Paid subscribers unlock the full library: 200+ decks, 10,000+ investors, AI Agent playbooks, and fresh high-signal content every week.

Here’s a good place to start:

🤖 AI Agent Templates
↳ The McKinsey Slide Playbook for Claude — how to build consulting-grade presentations in minutes.

📈 Startup Playbooks
↳ The AI Founders Playbook — lessons from Sequoia, OpenAI, Nvidia, and LangChain on building AI startups from zero.

🎯 The Deck Library
↳ 27 Most Promising AI Startup Pitch Decks Backed by Top Investors — the best-funded AI startups of 2025, slide by slide.

💰 Investor Database & Tools
↳ The Family Office List — 1,000 family offices actively writing cheques. Used by 5,000+ founders raising their round.

Not subscribed yet? Try it free for 7 days and unlock this pitch deck, plus our full premium library:

Get 7-Day Free Trial

We’re already 80,000+ founders and operators inside!

Let’s begin!

👇 Slide-by-slide Breakdown: What 1X is Really Building 🔥

Read more

Opus 5 shipped on July 24: near Fable 5 intelligence at half the price, with a real jump on numerical reasoning, tables, and financial modeling, the exact work a founder without a finance team keeps getting wrong.

At Startup School, Boris Cherny, the creator of Claude Code, said they deleted more than eighty percent of Claude Code’s system prompt for Opus 5 → His advice even to non-builders: every six months, delete your skills and config and see what happens. It might surprise you.

So this instead of being a pile of prompts, is six reusable Skills you build once and Opus 5 loads on its own.

📚 In this guide you’ll get copy-and-paste Skills to:

• Compare actuals to budget and prior month, explain every variance and list what to investigate

• Break down unit economics by channel and cohort, decompose gross margin and show which channel or cohort actually loses money

• Model a price change under different churn responses, give you the break-even churn and flag the segments that are underpriced

• Build a 12-month rolling forecast as an Excel file, run base, bear and bull scenarios, and tell you which single assumption moves your runway most

• Size the raise off your forecast, model the dilution across the cap table including the option pool, and check your target valuation against comparable rounds

• Build a live KPI and runway dashboard that pulls from your connected tools and generate the answer-first monthly investor update from the same numbers

Inside you will find:

1. How to Set Up Your Finance Skills So Opus 5 Runs on Your Real Numbers

2. How to Close the Month and See Where You Actually Stand

3. How to Find Where Your Unit Economics Leak Money

4. How Much Room You Actually Have to Raise Prices

5. How to Build a Forecast and Stress It Before You Bet on It

6. How to Size the Raise and Model the Dilution Before the First Meeting

7. How to Track the Numbers Live and Tell the Story Investors Trust

Launch your site faster than ever with Framer - now with Agents that make every part of your workflow effortless

Framer is the pro website builder for creators, teams, and businesses that care enough to get every detail right. Now with Agents that work alongside you across the full website workflow: designing new pages, managing your CMS, editing content, and adding SEO all without leaving the tool where the real site lives. Trusted by teams at companies like Miro and Perplexity.

Agents bring speed and scale; you bring taste, judgment, and control. Together you can:

• Build and iterate faster: generate sections, refine layout and styling, and ship updates in minutes.

• Scale content with confidence: update CMS collections, localize copy, and keep metadata consistent across pages.

• Stay production‑ready: reviewable changes, collaboration, and safeguards that help teams ship without handoffs.

Launch Your Site With Framer Today!

This issue is available exclusively to paid subscribers!

Paid subscribers unlock the full library: 200+ decks, 10,000+ investors, AI Agent playbooks, and fresh high-signal content every week.

Here’s a good place to start:

🤖 AI Agent Templates
↳ The McKinsey Slide Playbook for Claude — how to build consulting-grade presentations in minutes.

📈 Startup Playbooks
↳ The AI Founders Playbook — lessons from Sequoia, OpenAI, Nvidia, and LangChain on building AI startups from zero.

🎯 The Deck Library
↳ 27 Most Promising AI Startup Pitch Decks Backed by Top Investors — the best-funded AI startups of 2025, slide by slide.

💰 Investor Database & Tools
↳ The Family Office List — 1,000 family offices actively writing cheques. Used by 5,000+ founders raising their round.

Not subscribed yet? Try it free for 7 days and unlock this guide, plus our full premium library:

Get 7-Day Free Trial

We’re already 70,000+ founders and operators inside!

Let’s begin!

Read more

Last weekend, Y Combinator packed the Chase Center and handed the floor to the people actually building the future of AI.

Over 30,000 founders applied, from 120+ countries. Acceptance rate: under 1%. And the ones who got in? Jensen Huang told them flat out: “If you’re here, you’re gonna make it.”

Two days. Ten stages. Jensen Huang, Sam Altman, Boris Cherny, Jeff Dean, Waymo’s Dmitri Dolgov, and more. each one dropping the kind of insight most people pay years of tuition (or failed startups) to learn.

I sat through all of it, read the notes, the transcripts, and every founder’s live thread, so you don’t have to. Inside you’ll find the key takeaways from these speakers:

• Garry Tan (President, YC)

• Jensen Huang (Founder and CEO, NVIDIA)

• Sam Altman (Co-founder and CEO, OpenAI)

• Boris Cherny (Head of Claude Code, Anthropic)

• Jeff Dean (Chief Scientist, Google DeepMind)

• Dmitri Dolgov (Co-founder and CEO, Waymo)

• Alexandr Wang (Chief AI Officer at Meta)

• Patrick Collison (Co-founder and CEO, Stripe)

• Chelsea Finn (Co-founder, Physical Intelligence)

• Peter Steipete (OpenClaw)

Let’s get into it!

This issue is available exclusively to paid subscribers!

Paid subscribers unlock the full library: 200+ decks, 10,000+ investors, AI Agent playbooks, and fresh high-signal content every week.

Here’s a good place to start:

🤖 AI Agent Templates
↳ The McKinsey Slide Playbook for Claude — how to build consulting-grade presentations in minutes.

📈 Startup Playbooks
↳ The AI Founders Playbook — lessons from Sequoia, OpenAI, Nvidia, and LangChain on building AI startups from zero.

🎯 The Deck Library
↳ 27 Most Promising AI Startup Pitch Decks Backed by Top Investors — the best-funded AI startups of 2025, slide by slide.

💰 Investor Database & Tools
↳ The Family Office List — 1,000 family offices actively writing cheques. Used by 5,000+ founders raising their round.

Not subscribed yet? Try it free for 7 days and unlock this guide, plus our full premium library:

Get 7-Day Free Trial

We’re already 70,000+ founders and operators inside!

Let’s begin!

Read more

The strange part is that CEO Amjad Masad does not treat that number as a goal at all.

He warns that most AI companies posting growth like this are quietly dying underneath it, with churn approaching 100%, and that a headline like his own can hide a business that is falling apart.

So we went through hours of Masad’s interviews to find out what actually drove the run.

What follows is the sequence of moves behind the number: the layoff that turned into an all-in bet, the customer nobody else was chasing, the boring infrastructure that became the real moat and the honest way to read AI growth, including your own.

Inside you will find:

1. The Layoff That Became an All-In Bet on Agent

2. Build the Product That Gets Better When You Swap the Model

3. Chase the Billion Non-Engineers, Not the Developers

4. The Real Moat Is Two Years of Boring Infrastructure

5. Story Is a Growth Channel, Not a PR Function

6. How to Read AI Growth Honestly

7. What to Steal: The 6-Move Replit Playbook

This issue is available exclusively to paid subscribers!

Paid subscribers unlock the full library: 200+ decks, 10,000+ investors, AI Agent playbooks, and fresh high-signal content every week.

Here’s a good place to start:

🤖 AI Agent Templates
↳ The McKinsey Slide Playbook for Claude — how to build consulting-grade presentations in minutes.

📈 Startup Playbooks
↳ The AI Founders Playbook — lessons from Sequoia, OpenAI, Nvidia, and LangChain on building AI startups from zero.

🎯 The Deck Library
↳ 27 Most Promising AI Startup Pitch Decks Backed by Top Investors — the best-funded AI startups of 2025, slide by slide.

💰 Investor Database & Tools
↳ The Family Office List — 1,000 family offices actively writing cheques. Used by 5,000+ founders raising their round.

Not subscribed yet? Try it free for 7 days and unlock this guide, plus our full premium library:

Get 7-Day Free Trial

We’re already 70,000+ founders and operators inside!

Let’s begin!

Read more

ElevenLabs crossed $500M ARR in April 2026, up from $350M at the end of 2025.

Carles Reina built that revenue org from zero, as the company’s fourth employee and first investor: three years from nothing to past half a billion.

He recently laid out how he did it in a long, direct interview on Harry Stebbings’ 20VC, a set of rules specific enough to copy.

You can’t automate the engine, because the engine is people and some of his rules, like the 20x quota, stay there. What you can encode is the rest, the operational work:

• who to target

• which account is worth a call this week

• which deals are actually real

That is where Claude comes in, with Carles’s rules instead of generic best practices.

By the end you’ll have four Claude agents, each built from one of his rules:

• Concentrate on one proven ICP before scaling → the Concentration Agent

• Reach people humanly, not with transactional spray → the Timing Agent

• Land and expand → the Expansion Agent

• Public pipeline reviews and conservative forecasting → the Forecast Agent

Your reps keep the room, while your agents take the busywork.

Inside you will find:

1. Which Sales Rules You Can Actually Hand to Claude

2. Setting Up the Stack

3. How to Point Everything at One Winnable ICP

4. How to Get the Right Timing

5. How to Turn a First Small Contract Into a Whole-Account Plan

6. How to Strip the Optimism Out of Your Pipeline

7. Where the Agent Stops and the Rep Takes Over

1. Which Sales Rules You Can Actually Hand to Claude

Before you build anything, get honest about what is automatable and what is not. The fastest way to waste a month is to point an agent at work only a person can do.

Some of Carles’s rules are about people. They stay with you:

• 20x quota is a target you set and hold

• First contract in two weeks is a hiring and onboarding bar

• Reps in front of customers is a management choice

• Selling humanly, not transactionally is a principle, and it's his whole point: the moment outreach feels like AI spray, people ignore it

Build a sell more agent and you build nothing, because selling is the human part.

The other rules are really decision criteria and those can become software:

• Concentrate on one proven ICP is an instruction about where to point your list

• Reach people only when it's genuinely relevant, never with transactional volume

• Land and expand is a repeatable way to read an account and find the next door

• Conservative, public forecasting is a set of rules for deciding which deals are real

Each of those is a judgment applied over and over against fresh data, which is exactly what an agent is good at.

So the split you are working with:

STAYS HUMAN (guardrails, not agents)
- The 20x quota and the willingness to enforce it
- Selling humanly, not with transactional spray
- Reps in front of customers
- Hiring for hunters
- The room: negotiation, relationship, the final call on a deal

BECOMES AN AGENT (a decision rule applied to fresh data)
- The Concentration Agent  ← "concentrate on one proven ICP"
- The Timing Agent         ← "reach people only when it's relevant"
- The Expansion Agent      ← "land and expand through customer success"
- The Forecast Agent       ← "radically transparent pipeline reviews"

The rest of this guide builds those four agents, and each one takes its name and its job straight from one of Carles’s rules.

The “human rules” don’t disappear, they become the guardrails you write into the agents’ instructions, so everything Claude produces is filtered through the same discipline Carles applies himself.

The rest of the issue is available exclusively to paid subscribers!

Paid subscribers unlock the full library: 200+ decks, 10,000+ investors, AI Agent playbooks, and fresh high-signal content every week.

Here’s a good place to start:

🤖 AI Agent Templates
↳ The McKinsey Slide Playbook for Claude — how to build consulting-grade presentations in minutes.

📈 Startup Playbooks
↳ The AI Founders Playbook — lessons from Sequoia, OpenAI, Nvidia, and LangChain on building AI startups from zero.

🎯 The Deck Library
↳ 27 Most Promising AI Startup Pitch Decks Backed by Top Investors — the best-funded AI startups of 2025, slide by slide.

💰 Investor Database & Tools
↳ The Family Office List — 1,000 family offices actively writing cheques. Used by 5,000+ founders raising their round.

Not subscribed yet? Try it free for 7 days and unlock this guide, plus our full premium library:

Get 7-Day Free Trial

We’re already 70,000+ founders and operators inside!

Let’s begin!

Read more

💊 Before we start, today’s quick hits:

• My friend Jennifer just closed Passionfroot’s Series A. Go show her some love and say congrats here!

• LEARNING OF THE DAY: RETARDMAXXING, if Marc Andreessen is doing it, why shouldn’t you? Learn what it means here!

The YC Fall 2026 deadline is 5 days away

YC hands you $500K on founder-first terms and that’s almost the smallest part of it. You also get mentorship from people like Sam Altman or partners who helped scale Airbnb, Stripe, and Scale AI, plus access to a private network of alumni who’ve raised tens of billions in follow-on capital.

If you’re early, scrappy, and building something people want, it’s the most asymmetric bet you can make!

But getting in is harder. YC runs four batches a year now, so more people are applying and the bar keeps rising. Reviewers want more traction than they used to and the batch has gone almost entirely AI-native. A lot of the advice founders followed a few years ago just doesn’t clear that bar anymore.

What’s Inside the YC F26 Application Pack

📗 The 12 YC Application Principles That Actually Matter A tactical guide to what YC really looks for, backed by advice from Paul Graham, Dalton Caldwell and founders who got in successfully. It includes the exact frameworks, application breakdowns, and insider advice founders used to get into Dropbox, GitLab, Martin, and more.

🤖 3 Claude Skills That Review Your Application 3 free, open-source Claude skills that give you a brutally specific, source-grounded second read on your draft and prep you for the interview, without ever ghostwriting your application. Like having a YC alum look over your shoulder for as long as you need.

🍾 50+ Successful Applications (Dropbox, GitLab, Paystack...) A curated sheet with real YC applications from previous batches, including legacy names like Dropbox and GitLab, and newer ones like Paystack and KeyWords AI. Use them as references to model your own.

❓ Question-by-Question Answer Guide by Martin, Bolto, Shogun and Fluently’s Founders A doc-style walkthrough of the YC application, with practical advice on how to answer each question, straight from founders who got in (all in one place). No fluff. Just clear, tactical pointers for every field.

🙋 YC Summer 2026 “Request for Startups” A curated doc of the sectors and startup ideas YC is actively trying to fund right now. This is the most current list, since there’s no separate Fall 2026 one, so these asks still apply. Use it to sharpen your positioning or pitch a better idea entirely.

If you’re applying to YC F26, this is the most complete, battle-tested prep you’ll find. Everything you need to write a top-tier application in one place.

The next era of GTM runs on Attio.

It meets you where you work, compounds every customer signal into context, then acts on it so you can move fast, stay sharp, and scale without breaking. With Attio’s new workflows, run revenue agents across every account, so you can:

• Score every lead for fit and priority

• Catch renewals before they slip

• Get answers about your business in seconds

Loved by high-growth startups like Granola, Modal, and Railway, Attio is the agentic CRM for winning teams.

Try Now (For Free)

This issue is available exclusively to paid subscribers!

Paid subscribers unlock the full library: 200+ decks, 10,000+ investors, AI Agent playbooks, and fresh high-signal content every week.

Here’s a good place to start:

🤖 AI Agent Templates
↳ The McKinsey Slide Playbook for Claude — how to build consulting-grade presentations in minutes.

📈 Startup Playbooks
↳ The AI Founders Playbook — lessons from Sequoia, OpenAI, Nvidia, and LangChain on building AI startups from zero.

🎯 The Deck Library
↳ 27 Most Promising AI Startup Pitch Decks Backed by Top Investors — the best-funded AI startups of 2025, slide by slide.

💰 Investor Database & Tools
↳ The Family Office List — 1,000 family offices actively writing cheques. Used by 5,000+ founders raising their round.

Not subscribed yet? Try it free for 7 days and unlock this guide, plus our full premium library:

Get 7-Day Free Trial

We’re already 70,000+ founders and operators inside!

Let’s begin!

Read more

Written by Guillermo Flor · fact-checked against Kimi’s official documentation and current community reports on X · July 21, 2026

Kimi K3 is Moonshot AI’s flagship model for long-horizon coding, agentic knowledge work, reasoning, and multimodal tasks. Its API is OpenAI Chat Completions-compatible, it accepts text, images, and video, and it offers a 1,048,576-token context window. The model always reasons, and its preserved thinking must stay intact across multi-turn and tool-calling workflows.

The short verdict: K3 looks unusually capable when a task is large, visual, tool-heavy, or allowed to run for a while. It is less compelling when response latency is the main requirement. Early independent and community evidence also suggests that strong task completion does not automatically mean strong steerability or shell-error recovery. Give it explicit authority boundaries, acceptance tests, checkpoints, and a bounded agent loop.

Contents

1. K3 at a glance

2. What K3 is—and is not

3. Account setup and first API call

4. Parameters and reasoning

5. Preserved thinking and multi-turn state

6. Streaming

7. Images, video, and document files

8. Structured output and Partial Mode

9. Tool calling and agent loops

10. The 1M context window and automatic caching

11. Prompting K3 well

12. Pricing, rate limits, and cost control

13. Coding-agent integrations

14. What the community on X is seeing

15. Production checklist

16. Troubleshooting

17. Documentation inconsistencies to watch

18. Compact cheat sheet

Launch your site faster than ever with Framer - now with Agents that make every part of your workflow effortless

Framer is the pro website builder for creators, teams, and businesses that care enough to get every detail right. Now with Agents that work alongside you across the full website workflow: designing new pages, managing your CMS, editing content, and adding SEO all without leaving the tool where the real site lives. Trusted by teams at companies like Miro and Perplexity.

Agents bring speed and scale; you bring taste, judgment, and control. Together you can:

• Build and iterate faster: generate sections, refine layout and styling, and ship updates in minutes.

• Scale content with confidence: update CMS collections, localize copy, and keep metadata consistent across pages.

• Stay production‑ready: reviewable changes, collaboration, and safeguards that help teams ship without handoffs.

Launch Your Site With Framer Today!

Your subscription just got a serious upgrade.

Annual paid subscribers now get exclusive deals on the AI tools founders use every day, including:

• Framer: 1 year free

• MITO: 60% off for 6 months

• Gojiberry: 50% off for 6 months

• Cubic: 50% off your first month

• Omnia: 20% off

• Granola: 1 month free

• Slashy: 1 month free

• QuiverAI: 1 month free

• Throxy: 50% off for 3 months

These deals are worth more than $3,000!

Already on the annual plan? Access all your perks Here 👈

Not on the annual plan yet? There has never been a better time to upgrade!

Get 30% off forever

1. K3 at a glance

Read more

In June, Peter Steinberger (the creator of OpenClaw 🦞) said: “you shouldn’t be prompting coding agents anymore. You should be designing loops that prompt your agents.”

Founders who’d been treating their agent like a chatbot rebuilt it as a loop: propose, run, check, retry, and around again.

A month later (a couple of days ago):

The ground had moved again!

Here’s the part that actually matters: the loop that fixed your agent in June is the same loop setting you up to fail in a predictable way. A graph is what fixes that.

Picture the loop you probably already have running somewhere in your product: an agent drafts something, checks its own output against a rule, retries if it doesn’t pass. It looks disciplined, but it has a blind spot built into its shape.

The thing doing the checking is the same thing that did the work!

Your subscription just got a serious upgrade.

Annual paid subscribers now get exclusive deals on the AI tools founders use every day, including:

• Framer: 1 year free

• MITO: 60% off for 6 months

• Gojiberry: 50% off for 6 months

• Cubic: 50% off your first month

• Omnia: 20% off

• Granola: 1 month free

• Slashy: 1 month free

• QuiverAI: 1 month free

• Throxy: 50% off for 3 months

These deals are worth more than $3,000!

Already on the annual plan? Access all your perks Here 👈

Not on the annual plan yet? There has never been a better time to upgrade!

Get 30% off forever

Inside you will find:

1. Why the Loop That Got You Here Won’t Get You Further

2. What a Graph of Agents Actually Is

3. The Five Patterns Every Graph Is Built From

4. How to Tell If You Actually Need a Graph (or Just a Better Loop)

5. How to Build Your First Agent Graph in Claude

6. How to Keep a Graph From Becoming a Mess

This issue is available exclusively to paid subscribers!

Paid subscribers unlock the full library: 200+ decks, 10,000+ investors, AI Agent playbooks, and fresh high-signal content every week.

Get 7-Day Free Trial

We’re already 70,000+ founders and operators inside!

Let’s begin!

1. Why the Loop That Got You Here Won’t Get You Further

Strip any self-improving loop down to its skeleton and you find four moves repeating:

• pick something to control

• set a target

• measure the gap

• act to close it

• then go again

A thermostat is this in its purest form. So is the eval loop you probably already run on your agent: score an output, adjust the prompt or the retrieval when the score dips, ship, repeat.

The loop earns its popularity honestly. It’s the first thing that works, and watching a number respond to your adjustments is satisfying enough that it feels like the whole job.

The problem shows up later, and it’s structural: a loop can only see its own metric, so it will find every way to move that metric, including the ways that betray what the metric was supposed to represent!

Three blind spots come from this same shape and you’ve probably already met at least one of them:

1. The metric stops meaning what you think

Take a sales agent built to book more meetings. You wire a loop around it: measure meetings booked per week, adjust outreach copy and targeting when the number dips, ship, watch the following week. For a while it works exactly as advertised.

Then someone notices the meetings are with the wrong people, because the agent found that replying to anyone showing a flicker of interest books more meetings than replying only to qualified ones.

→ The loop didn’t fail; it optimized precisely what you told it to, on a target that had stopped meaning what you thought

2. The check is an echo

The check inside most loops is the agent checking itself, or a second agent built by the same hand, trained on the same assumptions, grading the first agent’s work. It looks like verification and behaves like verification. But two systems that share a blind spot will agree on that blind spot with total confidence.

3. The job outgrows the agent

A support-triage agent starts with a single job: read a ticket, tag it with a category. It works well, so you add one more instruction: also draft a suggested reply. Then another: also flag anything that looks like a refund request.

Then a few more, each one reasonable on its own, but each one bolted onto the same prompt, the same context window, the same agent. Six months in, that one agent is doing four unrelated jobs at once, and every edge-case rule you've added for job three has to coexist with the rules for jobs one, two, and four, whether or not they actually agree with each other.

None of this means loops are wrong. It means a single loop has a ceiling, and the ceiling isn’t about model quality!

2. What a Graph of Agents Actually Is

Read more

Instead of watching 2 hours of Netflix tonight, study this 40-minute masterclass from the founder of a $20B China AI company.

It's the clearest explanation I've seen of how Agent Swarms and AI systems actually work at scale.

I broke it down into the most actionable insights so you can apply them directly.

Here’s what you can learn today:

1. Why Kimi scales three dimensions at once instead of just adding tokens

2. Why token efficiency raises the ceiling on intelligence, not just the cost

3. Why long context is the capability that makes agents runnable for weeks

4. The 3 reward functions that stop agent swarms from collapsing or cheating

5. How Kimi fused all three dimensions into K2.5 (and why vision made text smarter)

6. The 24% token-efficiency jump hiding in Kimi’s next architecture

This issue is available exclusively to paid subscribers!

Paid subscribers unlock the full library: 200+ decks, 10,000+ investors, AI Agent playbooks, and fresh high-signal content every week.

Here’s a good place to start:

🤖 AI Agent Templates

↳ The McKinsey Slide Playbook for Claude — how to build consulting-grade presentations in minutes.

📈 Startup Playbooks

↳ The AI Founders Playbook — lessons from Sequoia, OpenAI, Nvidia, and LangChain on building AI startups from zero.

🎯 The Deck Library

↳ 27 Most Promising AI Startup Pitch Decks Backed by Top Investors — the best-funded AI startups of 2025, slide by slide.

💰 Investor Database & Tools

↳ The Family Office List — 1,000 family offices actively writing cheques. Used by 5,000+ founders raising their round.

Not subscribed yet? Try it free for 7 days and unlock this guide, plus our full premium library:

Get 7-Day Free Trial

We’re already 70,000+ founders and operators inside!

Let’s begin!

Read more

And this isn’t only about the moment someone asks. These are the documents the whole life of the startup runs on.

This is a curated collection of the 10 Templates and Databases every founder should have ready, always. They matter on the day you raise and they matter far more on all the ordinary days in between.

Here's what's inside:

1. Pitch collection: 90+ decks from the most successful startups ever, from Anthropic, OpenAI, and Perplexity to Airbnb and Revolut, plus fill-in templates from Sequoia and Creandum

2. Investors list: a database of 1,000+ angels, family offices, and VCs investing all around the world, filterable by check size, stage, and sector

3. Data room template: the complete Creandum structure investors expect the moment they start diligence.

4. Cap table plus SAFE dilution template: the template to check the right dilution and manage how your SAFEs dilute you as they convert

5. Valuation model: a defensible way to frame what your company is worth before the number comes up

6. Financial model template: the complete Creandum’s SaaS Model

7. Runway and burn template: how much you spend a month and how many months of cash that leaves

8. KPI and metrics dashboard calculator: the few numbers that describe the health of your business, computed right

9. Investor update template: the monthly note that keeps investors warm between rounds (by 500 Global)

10. The Startup Press Access Database: 500+ verified press touchpoints, journalists and publications that cover startups

And then there’s the part that makes this different from every other startup toolkit floating around.

The package comes with two things on top:

1. A prompt that interviews you inside Claude and turns your answers into a single company-context.md, which Claude then uses to fill any template from your real numbers

2. A Cowork setup: drop all ten documents in one folder, point Claude at it, and ask anything about your startup

Inside you will find:

1. The Five Documents That Get You Funded

2. The Three That Tell You If You’ll Survive the Year

3. The Two That Keep Investors and Press on Your Side

4. How to Fill Any of Them Without Starting From a Blank Page

5. How to Ask Your Files and Get One Answer

This issue is available exclusively to paid subscribers!

Paid subscribers unlock the full library: 200+ decks, 10,000+ investors, AI Agent playbooks, and fresh high-signal content every week.

Here’s a good place to start:

🤖 AI Agent Templates
↳ The McKinsey Slide Playbook for Claude — how to build consulting-grade presentations in minutes.

📈 Startup Playbooks
↳ The AI Founders Playbook — lessons from Sequoia, OpenAI, Nvidia, and LangChain on building AI startups from zero.

🎯 The Deck Library
↳ 27 Most Promising AI Startup Pitch Decks Backed by Top Investors — the best-funded AI startups of 2025, slide by slide.

💰 Investor Database & Tools
↳ The Family Office List — 1,000 family offices actively writing cheques. Used by 5,000+ founders raising their round.

Not subscribed yet? Try it free for 7 days and unlock this newsletter, plus our full premium library:

Get 7-Day Free Trial

We’re already 70,000+ founders and operators inside!

Let’s begin!

Read more

Finding the first customers is still brutal. And they’re the ones who matter most, because the earliest users are the only ones who:

• tell you whether the problem is real before you’ve spent a year building for no one

• give you the feedback that actually shapes the product, not the polite kind

• become the references and case studies that make the next ten easier to win

This week a repo went viral on X for taking a swing at exactly that.

It’s a Codex skill: you hand it your startup URL (or other info), it defines your ideal customer, reads public discussions for real pain and buying signals, qualifies each prospect with the source linked as evidence and drafts an opener.

So instead of trusting the demo, we ran an experiment. We made up a startup, an AI that chases down unpaid invoices for small businesses and freelancers, and gave the skill nothing to work with but that one sentence.

Then we let it loose in deep mode and watched what came back. And this is the final result (pretty crazy):

What comes back is a full report: an early-customer verdict, your ideal ICP, an evidence-backed shortlist of prospects with the public pain and buying signal behind each one, the patterns that repeat across them, a personalized opener per prospect, and a seven-day outreach plan."

In this issue we go deep on the parts that make it actually usable:

• how to install it, in Claude and not just Codex

• the six modes and when to reach for each

• what the tool did with just an idea, the real run, unfiltered

• how to schedule it so it works for you every day, dropping new prospects into an outreach queue you actually work

Inside you will find:

1. What the Skill Actually Does

2. How to Install It in Claude in Two Minutes

3. Which Mode to Use for Which Job

4. Putting the Skill to the Test

5. How to Turn It Into a Morning Radar

6. From Report to an Outreach Queue You’ll Actually Work

Launch your site faster than ever with Framer - now with Agents that make every part of your workflow effortless

Framer is the pro website builder for creators, teams, and businesses that care enough to get every detail right. Now with Agents that work alongside you across the full website workflow: designing new pages, managing your CMS, editing content, and adding SEO all without leaving the tool where the real site lives. Trusted by teams at companies like Miro and Perplexity.

Agents bring speed and scale; you bring taste, judgment, and control. Together you can:

• Build and iterate faster: generate sections, refine layout and styling, and ship updates in minutes.

• Scale content with confidence: update CMS collections, localize copy, and keep metadata consistent across pages.

• Stay production‑ready: reviewable changes, collaboration, and safeguards that help teams ship without handoffs.

Launch Your Site With Framer Today!

1. What the Skill Actually Does

Under the branding, it’s a disciplined research pipeline with a hard rule at the end.

1. It reads your product. It inspects the URL, the landing-page copy and the repository if you point it at one (or whatever describes your startup), then writes a brief: what you sell, the outcome you promise, who the user is, who actually pays, the urgent job, the current workaround, the trigger that makes someone start looking, and the disqualifiers. This brief is the filter for everything after it.

2. It searches five buckets, instead of hammering one query:

• Explicit demand: people literally asking for a tool or an alternative

• Pain: posts describing the problem as slow, manual, or infuriating

• Workaround: the spreadsheets and copy-paste rituals people build when no product fits

• Switching: cancellations, migration complaints, and missing-feature gripes aimed at a competitor

• Timing: public events like a launch, a hire, or an expansion that make your product relevant this month

3. It scores every prospect on five dimensions, and the weights tell you what it cares about:

• Pain strength (25%)

• Product fit (25%)

• Timing (20%)

• Public reachability (15%)

• Evidence quality (15%)

A prospect with no cited pain or timing signal never makes the primary shortlist, no matter how well it fits on paper.

4. It labels and drafts. Each prospect gets a stage, from high intent down to potential fit, plus one short opener that references the specific public context and nothing invented

This issue is available exclusively to paid subscribers!

Paid subscribers unlock the full library: 200+ decks, 10,000+ investors, AI Agent playbooks, and fresh high-signal content every week.

Here’s a good place to start:

🤖 AI Agent Templates
↳ The McKinsey Slide Playbook for Claude — how to build consulting-grade presentations in minutes.

📈 Startup Playbooks
↳ The AI Founders Playbook — lessons from Sequoia, OpenAI, Nvidia, and LangChain on building AI startups from zero.

🎯 The Deck Library
↳ 27 Most Promising AI Startup Pitch Decks Backed by Top Investors — the best-funded AI startups of 2025, slide by slide.

💰 Investor Database & Tools
↳ The Family Office List — 1,000 family offices actively writing cheques. Used by 5,000+ founders raising their round.

Not subscribed yet? Try it free for 7 days and unlock this guide, plus our full premium library:

Get 7-Day Free Trial

We’re already 70,000+ founders and operators inside!

2. How to Install It in Claude in Two Minutes

Read more

The thing between a solo founder and a real company was never talent. It was access. That access used to cost a salary. Now it costs a folder.

Here is the catch, and it’s why most founders miss it. Ask Claude “how do I validate my idea” and you get something that reads fine and helps you with nothing, because it doesn’t know your stage, your market, your constraints or the three decisions you made last week. Same question, generic answer.

Change the architecture and the answer changes completely.

A skill is a folder that carries the real framework, the diagnostic and the decision tree, and hands its output to the next skill so context compounds instead of resetting. Point a raw idea at the first one and it doesn’t lecture you on validation. It hands back a verdict, scored against a real checklist:

That is one skill, on one idea. Below is the full path the OS runs, from pressure-testing an idea to the cap table, and inside you’ll watch it run end to end on a real example. Each stage is a skill you build once and call when you reach it:

1. Idea Validation: decide in under an hour whether the idea holds, with Mom Test interview scripts and GO / PIVOT / KILL criteria instead of wishful thinking.

2. Business Model & Pricing: pick the revenue model, set the price, and check the unit economics (LTV, CAC, payback) before you commit a year to it.

3. Positioning: map the alternatives and claim the ground only you can hold, using April Dunford’s method rather than a tagline you invented in the shower.

4. Product: write the PRD, prioritize with RICE, and define the non-goals so scope creep doesn’t quietly kill the release.

5. Go-to-Market: go from a pre-launch waitlist to the first hundred customers with a launch sequence, not a hope that people show up.

6. Sales: run founder-led selling, qualify with MEDDIC or BANT depending on the deal, and work outreach sequences that actually get replies.

7. Growth & Analytics: instrument AARRR, pick a North Star, read your retention and churn curves, and size an A/B test so the result means something.

8. Marketing & Brand: put distribution ahead of creation, build pillar content, and run an SEO play that compounds instead of decays.

9. Fundraising: build a Sequoia/YC-style deck and work the outreach hierarchy that gets warm intros instead of cold silence.

10. Finance, Legal & Cap Table: track burn and runway, choose the entity, and handle SAFEs and the cap table without a surprise you can’t undo.

    

Inside you will find:

1. How the Founder OS Works (The Setup)

2. How You Actually Run It

3. The 10 Skills That Run Your Solo Business

4. How to Keep the OS Alive and Encode Your Own Unfair Advantage

This issue is available exclusively to paid subscribers!

Paid subscribers unlock the full library: 200+ decks, 10,000+ investors, AI Agent playbooks, and fresh high-signal content every week.

Here’s a good place to start:

🤖 AI Agent Templates
↳ The McKinsey Slide Playbook for Claude — how to build consulting-grade presentations in minutes.

📈 Startup Playbooks
↳ The AI Founders Playbook — lessons from Sequoia, OpenAI, Nvidia, and LangChain on building AI startups from zero.

🎯 The Deck Library
↳ 27 Most Promising AI Startup Pitch Decks Backed by Top Investors — the best-funded AI startups of 2025, slide by slide.

💰 Investor Database & Tools
↳ The Family Office List — 1,000 family offices actively writing cheques. Used by 5,000+ founders raising their round.

Not subscribed yet? Try it free for 7 days and unlock this guide, plus our full premium library:

Get 7-Day Free Trial

We’re already 70,000+ founders and operators inside!

Let’s begin!

Read more

Today I’m sharing a curated list of the angel investors behind many of the biggest AI funding rounds of 2026.

We started from the most notable AI companies that raised capital this year, including and for each investor, we included the essentials you’d actually want if you were researching them or planning outreach:

• Name

• Role (founder, operator, CEO, angel investor, etc.)

• LinkedIn profile

• Company backed

So instead of just knowing who these angels are, you can also quickly understand why they matter, what kind of operator or founder they are and which breakout AI companies they’ve backed.

If you’re a founder, operator, or investor, this gives you a much clearer view of the people showing up behind the most important AI companies being funded right now.

Launch your site faster than ever with Framer - now with Agents that make every part of your workflow effortless

Framer is the pro website builder for creators, teams, and businesses that care enough to get every detail right. Now with Agents that work alongside you across the full website workflow: designing new pages, managing your CMS, editing content, and adding SEO all without leaving the tool where the real site lives. Trusted by teams at companies like Miro and Perplexity.

Agents bring speed and scale; you bring taste, judgment, and control. Together you can:

• Build and iterate faster: generate sections, refine layout and styling, and ship updates in minutes.

• Scale content with confidence: update CMS collections, localize copy, and keep metadata consistent across pages.

• Stay production‑ready: reviewable changes, collaboration, and safeguards that help teams ship without handoffs.

Launch Your Site With Framer Today!

Let’s dig in! Below are the companies covered in the database:

ElevenLabs

AI audio platform building models for speech generation, dubbing, sound effects, music and conversational voice agents.

Funding: Raised a $500M Series D at an $11B valuation in February 2026, led by Sequoia, with participation from a16z and ICONIQ. The round brought total funding to $781M.

Angels in the database: 19

Lovable

AI software-development platform that enables users to build and deploy applications through natural-language prompts.

Funding: Raised a $330M Series B at a $6.6B valuation in December 2025, led by CapitalG and Menlo Ventures, with backing from Accel, Khosla Ventures, DST Global, Salesforce Ventures and others. It has raised approximately $552.5M in total. Lovable is included as a particularly relevant exception to the 2026 funding criterion.

Angels in the database: 18

Basis

AI agent platform designed to automate complex accounting, tax and audit workflows.

Funding: Raised a $100M Series B at a $1.15B valuation in February 2026, led by Accel, with participation from GV and existing investors. Total funding is approximately $138M.

Angels in the database: 12

Decagon

Enterprise platform for building conversational AI agents that automate customer-service operations.

Funding: Raised a $250M Series D at a $4.5B valuation in January 2026, led by Coatue and Index Ventures. Other backers include a16z, Accel, Bain Capital Ventures, Forerunner and Ribbit Capital. Total funding is approximately $481M.

Angels in the database: 9

Runway

AI research and creative platform developing generative video models and general-purpose world models.

Funding: Raised a $315M Series E at a reported $5.3B valuation in February 2026, led by General Atlantic, with participation from NVIDIA, Adobe Ventures, AMD Ventures, Fidelity and others. Total funding is approximately $860M.

Angels in the database: 8

TwelveLabs

Video-intelligence company building multimodal models capable of understanding, searching and reasoning over video.

Funding: Raised a $100M Series B in July 2026, co-led by NEA and NAVER Ventures, with Amazon among its strategic backers. Public databases place total funding at approximately $150M–$210M.

Angels in the database: 8

Cognition

Developer of Devin, an autonomous AI software engineer capable of handling complex development tasks.

Funding: Raised more than $1B at a $26B post-money valuation in May 2026, co-led by Lux Capital, General Catalyst and 8VC, with participation from Founders Fund, Ribbit Capital and Atreides. Reported total funding now exceeds $2.5B.

Angels in the database: 7

Norm AI

Legal and compliance platform using AI agents to interpret regulations and execute regulated workflows.

Funding: Raised a $120M Series C at a $1.2B valuation in July 2026, led by Khosla Ventures. Total funding is approximately $256M.

Angels in the database: 5

OpenEvidence

Clinical AI platform that gives physicians evidence-based answers sourced from medical literature and clinical data.

Funding: Raised a $250M round at a $12B valuation in January 2026, co-led by Thrive Capital and DST Global. The company said it had raised approximately $700M over the previous 12 months. Other backers include GV, NVIDIA, Kleiner Perkins and Mayo Clinic.

Angels in the database: 4

Factory

AI software-engineering platform building autonomous “Droids” for enterprise development teams.

Funding: Raised a $150M Series C at a $1.5B valuation in April 2026, led by Khosla Ventures, with participation from Sequoia, Blackstone, Insight Partners, NEA and others. Total funding is approximately $220M.

Angels in the database: 3

Harvey

AI operating system for law firms and professional-services teams, covering research, due diligence, compliance and other legal workflows.

Funding: Raised $200M at an $11B valuation in March 2026, co-led by GIC and Sequoia. Other backers include a16z, Coatue, Kleiner Perkins and Conviction. Harvey has raised more than $1B in total.

Angels in the database: 3

Odyssey

AI research company developing world models that can understand and simulate dynamic physical environments.

Funding: Raised a $310M Series B at a $1.45B valuation in June 2026, led by Natural Capital, with participation from Amazon, AMD Ventures, GV, EQT and In-Q-Tel. Total funding is approximately $337M.

Angels in the database: 2

Replit

AI-native software-creation platform that allows users to build, deploy and manage applications from natural-language instructions.

Funding: Raised a $400M Series D at a $9B valuation in March 2026. Public funding estimates range from approximately $852M to $922M, with PitchBook reporting about $878M.

Angels in the database: 2

Suno

Generative AI platform that enables users to create complete songs and other music from text prompts.

Funding: Raised a $400M Series D at a $5.4B valuation in June 2026, following a $250M Series C in late 2025. The company has raised approximately $775M across its disclosed rounds, backed by firms including Menlo Ventures, Lightspeed, Matrix, IVP and Bond.

Angels in the database: 2

Flourish

AI research company combining neuroscience and machine learning to develop more efficient, continuously learning AI architectures.

Funding: Launched with approximately $500M in funding at a reported $2.5B valuation in 2026, with backing from Bezos Expeditions, Lux Capital, GV and Catalio.

Angels in the database: 1

Garner Health

Healthcare-navigation platform using claims data to identify high-quality doctors and guide employees toward better care.

Funding: Raised a $118M Series D at a $1.35B valuation in February 2026, led by Kleiner Perkins, with participation from Redpoint, Maverick, Kaiser Permanente Ventures, Mercy and Plus Capital. Total funding is approximately $200M.

Angels in the database: 1

General Intuition

Frontier AI lab training agents on gameplay data to develop spatial reasoning and real-world intelligence.

Funding: Raised a $320M Series A at a $2.3B valuation in June 2026, led by Khosla Ventures, with participation from General Catalyst, Hedosophia, Innovation Endeavors and other backers. Total disclosed funding is approximately $454M.

Angels in the database: 1

Modus

AI-native audit technology platform and holding company investing in and partnering with accounting firms.

Funding: Announced $85M across its seed and Series A financing in April 2026, led by Lightspeed Venture Partners, with participation from Comma Capital and Garry Tan. This represents its total publicly disclosed funding.

Angels in the database: 1

This database is available exclusively to paid subscribers!

Along with it, you’ll unlock our full library: 200+ startup decks, 10,000+ investors, AI Agent playbooks, and fresh high-signal content every week.

Not subscribed yet? Try it free for 7 days and unlock the database, plus our full premium library:

Get 7-Day Free Trial

We’re already 70,000+ founders and operators inside!

Access the list here 👇

Read more

Tom Blomfield, the YC partner who founded Monzo, calls that a broken way of thinking, “like taking the old way of working and adding like a more powerful engine onto it.”

The real move is rebuilding the company itself as a set of recursive self-improving loops and YC is already running three of them internally.

Blomfield delivered a chat to a room of founders, building on ideas from Diana Hu, Garry Tan and a Jack Dorsey thread.

I pulled it apart into the exact framework, the live examples, and the org changes, so you can start building your company in this shape this week.

In this guide you’ll find:

1. Why the Copilot Framing Is a Trap

2. The 5-Layer Loop Behind a Self-Improving Company

3. How YC’s Agent Fixes Failed Queries Overnight

4. Why Tokens Replace Headcount as the Constraint

5. How to Make Your Company Legible to AI

6. Why Data Is Precious and Software Is Ephemeral

7. What to Steal This Week

This issue is available exclusively to paid subscribers!

Paid subscribers unlock the full library: 200+ decks, 10,000+ investors, AI Agent playbooks, and fresh high-signal content every week.

Here’s a good place to start:

🤖 AI Agent Templates
↳ The McKinsey Slide Playbook for Claude — how to build consulting-grade presentations in minutes.

📈 Startup Playbooks
↳ The AI Founders Playbook — lessons from Sequoia, OpenAI, Nvidia, and LangChain on building AI startups from zero.

🎯 The Deck Library
↳ 27 Most Promising AI Startup Pitch Decks Backed by Top Investors — the best-funded AI startups of 2025, slide by slide.

💰 Investor Database & Tools
↳ The Family Office List — 1,000 family offices actively writing cheques. Used by 5,000+ founders raising their round.

Not subscribed yet? Try it free for 7 days and unlock this guide, plus our full premium library:

Get 7-Day Free Trial

We’re already 70,000+ founders and operators inside!

Let’s begin!

Read more

The company reports going from $0 to over $100M in annualized revenue in roughly 9 months while selling an integrated RL post-training stack. They then raised a $130M Series A led by Radical Ventures with NVIDIA, Intel Capital, and Dell Capital.

We broke down the exact wedge, packaging, and customer-proof sequence behind that run so you can steal the playbook, not just admire the number.

In this guide you’ll find:

1. What Prime Intellect actually sells: the Open Superintelligence Stack, layer by layer

2. The open-source wedge: how the Environments Hub reached 1,000+ environments before Lab launched

3. The positioning bet that unlocked demand: how ‘RL breaks open the frontier’ won teams away from closed-lab APIs

4. The 4 growth levers that drove 6,000+ customers and $100M ARR in under a year

5. The named-customer proof machine (Ramp, Arcee, Zapier) and how it replaced a sales team

6. The $130M Series A: who backed it, at what traction, and the angel list

7. 5 lessons to steal from Prime Intellect’s 9-month run

👋 Just a quick note before you continue!

From now on, as a premium subscriber, you’ll get access not just to the best resources on fundraising and growth (160+ Playbooks, 65+ Success Stories, 30+ Pitch Deck Collections, 25+ Investor Resources), but also:

Annual paid subscribers now unlock exclusive deals on the AI tools founders actually use every day.

Here’s what’s live on the perks page right now:

• Framer: 1 year free

• MITO: 60% off for 6 months

• Gojiberry: 50% off for 6 months

• Cubic: 50% off your first month

• Omnia: 20% off

• Granola: 1 month free

• Slashy: 1 month free

• QuiverAI: 1 month free

• Throxy: 50% off for 3 months

These deals are worth more than $3,000.

Get 30% off forever

Read more

Elon Musk does not outperform entire industries because he works longer hours, hires more people or has access to better technology.

He operates his companies using a distinct system for questioning assumptions, removing complexity, accelerating learning, identifying exceptional talent and concentrating resources on the biggest constraint.

To understand how that system actually works, I studied three of the clearest breakdowns of Musk’s operating methods:

David Senra’s 110-minute How Elon Thinks conversation with Eric Jorgenson, who spent five years and thousands of hours studying Musk to write The Book of Elon.

Relentless’s 102-minute conversation with Sequoia partner Shaun Maguire, Why Elon Outcompetes Everyone, which explores Musk’s talent selection, capital allocation, risk tolerance and ability to compound knowledge faster than his competitors.

And Andrej Karpathy’s talk at Sequoia AI Ascent, where the former Tesla AI leader explained what it was actually like to operate inside one of Musk’s companies—and how small, highly technical teams, extreme ownership and direct access to the underlying work shaped Tesla’s execution.

Together, they reveal something more useful than another collection of Elon stories.

They reveal the Elon Operating System: the repeatable principles Musk uses to build companies that move faster, learn faster and achieve outcomes that initially appear physically or economically impossible.

In this guide, you’ll learn:

1. The five-step build algorithm, in the exact order Musk runs it

2. How Musk engineers his companies to fail on purpose

3. How Musk identifies talent and creates extreme accountability

4. Why Musk keeps teams small, technical and close to the work

5. The mission-and-pain layer beneath the operating system

6. How Musk allocates capital and attacks constraints

7. The portable Elon OS checklist

👋 Just a quick note before you continue!

From now on, as a premium subscriber, you’ll get access not just to the best resources on fundraising and growth (160+ Playbooks, 65+ Success Stories, 30+ Pitch Deck Collections, 25+ Investor Resources), but also:

Annual paid subscribers now unlock exclusive deals on the AI tools founders actually use every day.

Here’s what’s live on the perks page right now:

• Framer: 1 year free

• MITO: 60% off for 6 months

• Gojiberry: 50% off for 6 months

• Cubic: 50% off your first month

• Omnia: 20% off

• Granola: 1 month free

• Slashy: 1 month free

• QuiverAI: 1 month free

• Throxy: 50% off for 3 months

These deals are worth more than $3,000.

Get 30% off forever

Here’s a good place to start:

🤖 AI Agent Templates

↳ The McKinsey Slide Playbook for Claude — how to build consulting-grade presentations in minutes.

📈 Startup Playbooks

↳ The AI Founders Playbook — lessons from Sequoia, OpenAI, Nvidia, and LangChain on building AI startups from zero.

🎯 The Deck Library

↳ 27 Most Promising AI Startup Pitch Decks Backed by Top Investors — the best-funded AI startups of 2025, slide by slide.

💰 Investor Database & Tools

↳ The Family Office List — 1,000 family offices actively writing cheques. Used by 5,000+ founders raising their round.

The rest of the article is only available to premium subscribers.

1. The 5-step build algorithm, in the exact order Musk runs it (and why running it backwards nearly killed Tesla and Starlink)

Read more

Three weeks after SpaceX signed the $60B agreement to acquire Cursor, the two companies shipped their first product together.

“Announcing Grok 4.5, our first model trained specifically for coding and agents. It was trained with Cursor and offers frontier intelligence at leading speeds and cost efficiency.”

Two numbers tell you why this matters:

• $2 / $6 per million tokens (input/output), against the $10/$50 of Claude Fable 5, the current benchmark leader

• Trillions of tokens of real Cursor usage in the training data. Nobody else has that dataset

But here is the real point for you. Models rotate every six weeks. The place where you work with them is consolidating and that place is Cursor: no longer a code editor but a desktop app, a web app, an iPhone app, a CLI, cloud agents, a code reviewer, and now co-author of its own frontier model. Half of the Fortune 500 pays for it.

And yet most founders still use it like an expensive autocomplete, or skip it entirely because “I’m not technical enough.”

Which means this news just made the gap bigger: the founders who operate Cursor properly are compounding every week, and everyone else is chatting with an expensive chatbot.

This guide closes that gap. Everything from zero: how to set it up, which plan to pick, how to run it next to Claude Code and Codex and the part that alone is worth the read: five founder workflows only Cursor makes possible.

So this edition is two things in one:

1. The News Decoded

2. The Founder’s Beginner Guide to Cursor

Inside you will find:

🟧 What Grok 4.5 is, and the honest benchmarks

🟧 The seven surfaces of Cursor in 2026

🟧 How to start from zero, with no code at all

🟧 Cursor vs Claude Code vs Codex and how to run them together

🟧 The full setup

🟧 The first-session script that builds trust before you build

🟧 The X-ray: what your developers actually shipped this week

🟧 Five founder workflows only Cursor makes possible

Let’s begin!

1. What Grok 4.5 Actually Is (and Why It Was Trained With Cursor)

The short version: Grok 4.5 is a mixture-of-experts model trained by SpaceXAI (the SpaceX and xAI merger entity) together with Cursor’s model team, on Colossus infrastructure, across tens of thousands of NVIDIA GB300 GPUs.

The collaboration started in April with a compute partnership. The $60B acquisition came in June, days after the SpaceX IPO.

The Benchmarks

Grok 4.5 is not the smartest model available. It is the most efficient one:

• SWE Marathon (can the agent work on a project for hours without losing the thread?): 29.0%, first place. Sounds low, but nobody has ever scored higher

• Terminal Bench 2.1 (can it operate a terminal on its own: run commands, install things, recover from errors?): 83.3%, tied with Claude Fable 5 (84.3%) and GPT-5.5 (83.4%)

• DeepSWE 1.0 (real bugs and features from GitHub repos): 62.0%, behind Fable 5 (66.1%) and GPT-5.5 (64.3%), ahead of Opus 4.8 (55.8%)

• SWE-Bench Pro (same idea, hardest problems): 64.7%, well behind Fable 5 (80.3%). On raw intelligence, Anthropic still wins clearly

The number that changes behavior is efficiency: Grok 4.5 solves a SWE-Bench Pro task in ~16,000 output tokens on average. Opus 4.8 needs 67,000. That is 4.2x fewer tokens, at 80 tokens per second, at a fraction of the price.

Where You Can Use It

• SpaceXAI API: $2/$6 per million tokens, fast variant at $4/$18. If your product calls an LLM, you can test it tomorrow without installing anything

• Grok Build (xAI’s CLI): it is the default model there, free for a limited time

• Inside Cursor: available today on every plan, across desktop, web, iOS, CLI and SDK, with double usage for the first week. Not the default, not exclusive: Composer 2.5 stays

• EU availability: expected mid-July

👇 Now let’s go ahead with the complete founder’s guide to using Cursor

This second part is available exclusively to paid subscribers!

Paid subscribers unlock the full library: 200+ decks, 10,000+ investors, AI Agent playbooks, and fresh high-signal content every week.

Here’s a good place to start:

🤖 AI Agent Templates
↳ The McKinsey Slide Playbook for Claude — how to build consulting-grade presentations in minutes.

📈 Startup Playbooks
↳ The AI Founders Playbook — lessons from Sequoia, OpenAI, Nvidia, and LangChain on building AI startups from zero.

🎯 The Deck Library
↳ 27 Most Promising AI Startup Pitch Decks Backed by Top Investors — the best-funded AI startups of 2025, slide by slide.

💰 Investor Database & Tools
↳ The Family Office List — 1,000 family offices actively writing cheques. Used by 5,000+ founders raising their round.

Not subscribed yet? Try it free for 7 days and unlock this guide, plus our full premium library:

Get 7-Day Free Trial

We’re already 70,000+ founders and operators inside!

2. What Can You Actually Do With Cursor

Read more

This morning Sifted reported that Lovable is in talks to raise $300 million at a $13.2 billion post-money valuation, with Menlo Ventures set to lead the round.

If it closes at those terms, the company will have doubled its price tag in roughly six months!

Sit with the timeline for a second, because the pace is the whole story:

• Oct 2024 → raises $7.5M (Hummingbird Ventures, byFounders)

• Feb 2025 → raises $15M (Creandum, Visionaries Club)

• Jul 2025 → becomes a unicorn, $200M Series A at $1.8B (Accel, 20VC, Creandum, Hummingbird, byFounders, Visionaries Club, Evantic)

• Dec 2025 → raises $330M at $6.6B (Google, Menlo Ventures, NVIDIA, Salesforce Ventures, Databricks Ventures, Atlassian Ventures, HubSpot Ventures, T.Capital, Khosla Ventures, DST Global, EQT Growth, Kinship Ventures, Accel, Creandum, Evantic)

• Jul 2026 → reported talks at $13.2B

That is five financings in about twenty months, each at a higher multiple than the last, with a founder who at one point simply stopped answering the term sheets landing in his inbox at a $4 billion valuation because he didn’t feel like raising yet.

Most founders experience fundraising as a thing that happens to them. Lovable built a machine that made investors show up before it asked, compete once they were there and accept terms that got worse for them with every passing quarter.

Inside you will find:

🟧 The demand machine that took Lovable from a $7.5M pre-seed to a reported $13.2B in twenty months

🟧 The full details of all five rounds and why each fund and angel was chosen

🟧 The angel-first move that buys competence and distribution, not just a check

🟧 The public ARR cadence that made investors chase before Lovable ever asked

🟧 What Osika did when unsolicited $4 billion offers hit his inbox

🟧 How the product closed a $200M round before the pitch meeting even started

🟧 The investor-sequencing map for who to raise from at each stage

🟧 The six channels that put your growth in front of the people who write the checks

The next era of GTM runs on Attio.

It meets you where you work, compounds every customer signal into context, then acts on it so you can move fast, stay sharp, and scale without breaking. With Attio’s new workflows, run revenue agents across every account, so you can:

• Score every lead for fit and priority

• Catch renewals before they slip

• Get answers about your business in seconds

Loved by high-growth startups like Granola, Modal, and Railway, Attio is the agentic CRM for winning teams.

Try Now (For Free)

This issue is available exclusively to paid subscribers!

Paid subscribers unlock the full library: 200+ decks, 10,000+ investors, AI Agent playbooks and fresh high-signal content every week.

Here’s a good place to start for fundraising:

🎯 The Deck Library
↳ 27 Most Promising AI Startup Pitch Decks Backed by Top Investors
↳ How ElevenLabs Grew From $0 to $200M ARR in 2 Years + Their Pre-Seed Deck
↳ 70 Pitch Decks That Raised Over $1 Billion in 2024
↳ Anthropic Pitch Deck: This 10 Slides Predicted a $350,000,000,000 Company

💰 Investor Database & Tools
↳ The Family Office List
↳ How to Get Into YC: The Full F25 Application Pack
↳ The 10 Excel Templates Every SaaS & AI Founder Needs to Build, Survive, and Scale
↳ Creandum fundraising Toolbox, from Seed to Series A

Not subscribed yet? Try it free for 7 days and unlock this guide, plus our full premium library:

Get 7-Day Free Trial

We’re already 70,000+ founders and operators inside!

Let’s begin!

Read more