A change. Starting April 1st, 2026, I’m pausing all paid subscriptions on Substack. If you’re a paying subscriber, you won’t be charged again (if technically all goes well :)).

You’ll still get a full year of paywalled content through April 1st, 2027. After that, the newsletter goes fully free for the first time since I started writing it in 2018, on a biweekly schedule.

In today’s edition, among other things:

• The Org Chart Has a Half-Life

• The AI Roll-up Thesis

• The Startup Playbook Is a Trap

• Securing AI Agents: The Defining Cybersecurity Challenge of 2026

• Founder-Led Sales Playbook: Repeatable Motion

• Corporate Venture Capital: A User Guide

Onwards!

The Org Chart Has a Half-Life

The org chart itself is becoming a temporary structure, compressed and redrawn faster than any management theory can keep up with.

The numbers from Foundation Capital’s research are crazy. One company is shrinking a 120-person engineering team to 25. Another running 30+ microservices has gone from 0.75 engineers per service to a projected 0.1, meaning a single engineer overseeing what used to require eight. A company with a 1:6 expert-to-generalist ratio today is targeting 1:25 within twelve months and eventually 1:100.

Those ratios describe a different kind of organization, one where traditional role boundaries dissolve rather than stretch.

Foundation Capital uses the electricity-and-factories analogy: most companies are still running AI on organizational architectures designed for a pre-AI world. Faster individuals don’t automatically produce a faster org when the handoffs, approval chains, and role boundaries stay the same. The taxonomy of what comes next is where it gets useful.

They propose four human roles that survive the compression: accountability officers (executives who own outcomes and sign filings), systems architects (people who design human-agent workflows), relationship experts (sales, account management, culture builders), and validators (humans who review and approve agent output). Nine traditional functions collapse into three: R&D, GTM, and G&A.

The validator role is the most interesting because it comes with a built-in paradox. Foundation Capital expects validator demand to follow a bell curve: ramping now, peaking in two to four years as agent deployments scale, then declining as systems accumulate enough data to self-correct. But the people qualified to validate are experts who built that expertise by doing the work themselves. If agents handle all the junior analyst work and first-draft code, the class of 2035 never gets those reps.

The validator pool is a one-generation asset unless we deliberately replenish it.

Every AI transformation roadmap assumes a supply of qualified validators that the transformation itself is depleting. The mandate at one 1,000-person company that every PM and designer must ship code using AI tools (”My First Pull Request”) is a partial answer. But building an app with Cursor is not the same as understanding why a system fails at 3 a.m.

The piece also maps a useful 2x2: product vs. service companies, digital vs. physical delivery. Digital services (consulting firms, BPOs, agencies) face the most immediate pressure because the product is the work itself, and agents are doing it. Digital product companies compress headcount but keep human oversight in the loop. Physical services are furthest from full automation but have the clearest opportunity for orchestration around scheduling and routing.

The startup opportunity that stands out is agent-native tooling. Foundation Capital notes that an MCP tool schema can consume roughly 55,000 tokens just to load, while the equivalent CLI command costs around 200. Bolting MCP onto Jira still routes agents through data models built for human cognition. The companies that build tools where agents are the primary user, with human oversight as a feature rather than the default interface, will outperform on raw throughput.

The companies in this research are restructuring roles, collapsing functions, and rethinking what “headcount” even means.

Source: The Great Reorg: A Human’s Guide.

The AI Roll-up Thesis

A “roll-up” is an old private equity playbook: buy a bunch of small businesses in the same industry, combine them under one roof, and extract value through scale. Consolidate back-office, centralize purchasing, apply one management layer across many locations. PE firms have done this for decades with dental practices and veterinary clinics.

An AI roll-up adds a different lever. You acquire small service businesses, the kind where most of the cost structure is people doing knowledge work: accounting firms, property managers, IT support shops, insurance brokers, logistics coordinators. Then you apply AI to automate a meaningful chunk of that work. A bookkeeping firm running on 15% EBITDA margins might get pushed toward 40-50% if AI handles the routine data entry, reconciliation, and reporting that currently requires junior staff. The difference from a traditional roll-up: you’re transforming the cost structure of the acquired business itself, not just consolidating overhead across a portfolio.

The model went from obscure to unavoidable in 2025. Google Trends for “AI roll-up” went from near-zero to peak interest in under a year, and over 150 companies are now on the AI Roll-up Nexus tracker. The economics are straightforward: services businesses trade at 4-6x EBITDA, so you buy cheaply, double margins through AI automation, and sell the combined entity at a higher multiple.

But the capital to prove the model arrives after the proof is already done.

Read more

In today’s edition, among other things:

• The Best People Don’t Look Like the Best People

• The FarmVille Problem in Enterprise AI

• The Space Economy’s Shipping Container Moment

• When the Factory Is the Moat

• Speed Is a Safety Problem

Onwards!

Read more

Today - a more thorough analysis, but worth reading.

In today’s edition:

• Wright’s Law Is Eating Everything

• The Intelligence Cost Collapse

• Benchmarks Are Saturating (And That’s the Point)

• Agents Are the Demand Multiplier

• $300 Bil…

Read more

In today’s edition, among other things:

• The AI Primitives Mess

• The $400 Million Bottleneck

• Too Valuable to Regulate

• The $5 Trillion Herd

• It’s time for agentic video editing

• The Rise of Agentic Commerce

Onwards!

Read more

In today’s edition, among other things:

• The 700-Agent Company

• The Case for Decision Traces

• The Web Is Tapped Out. Expert Data Is the New Oil.

• The Math Behind “Building in Public”

• The Marketplace Extinction Map

• Vertical AI Isn’t Competing for Your Software Budget

Onwards!

The 700-Agent Company

Notion has 1,000 employees and 700 AI agents. That ratio tells you more about the next five years of company building than most fundraising decks will.

Ivan Zhao’s recent post frames AI through industrial metaphors: steel, steam, bicycles, cars. The historical parallels are nice, but the concrete data point is what matters. A $10B company is running with roughly 0.7 agents per employee, and Zhao calls this “baby steps.”

Alongside our 1,000 employees, more than 700 agents now handle repetitive work. They take meeting notes and answer questions to synthesize tribal knowledge. They field IT requests and log customer feedback. They help new hires onboard with employee benefits.

The list is instructive: organizational sludge that scales linearly with headcount. Onboarding, IT tickets, meeting notes, status updates. Notion automated the communication overhead that makes companies slow, not their core product work.

AI is absorbing the coordination tax.

Zhao identifies two blockers for broader knowledge work automation: context fragmentation and verifiability. Code agents work because everything lives in one place (the IDE, the repo) and outputs can be tested. Knowledge work is scattered across Slack, docs, dashboards, and someone’s memory. You can verify if code compiles. You can’t easily verify if a strategy memo is good.

Today, humans are the glue, stitching all that together with copy-paste and switching between browser tabs. Until that context is consolidated, agents will stay stuck in narrow use-cases.

Notion cares about being the single context layer because if you’re building agents, the company that owns consolidated context wins. Fragmented SaaS stacks become a liability, not a feature.

Zhao’s critique of current AI adoption:

We’re still in the “swap out the waterwheel” phase. AI chatbots bolted onto existing tools. We haven’t reimagined what organizations look like when the old constraints dissolve.

The waterwheel metaphor comes from early steam adoption. Factory owners replaced water wheels with steam engines but kept everything else identical. Productivity gains were minimal. The breakthrough came when they redesigned entire factories around the new power source.

Most companies are doing the AI equivalent of swapping the waterwheel. Same processes, same meetings, same approval chains, plus a chatbot.

The operational implications:

1. Audit your coordination overhead. What percentage of employee time goes to syncing, updating, reporting, meeting? That’s your agent opportunity space.

2. Consolidate before you automate. Agents struggle with fragmented context. If your work lives in 15 different tools, you’ll get 15 mediocre point solutions.

3. Design for supervision, not intervention. Zhao references the 1865 Red Flag Act, which required someone to walk in front of every car waving a flag. Human-in-the-loop often means human-as-bottleneck. The goal is humans supervising from a leveraged point.

What would change my view: evidence that highly fragmented tool stacks with narrow AI solutions outperform consolidated platforms with deep integration. I haven’t seen it yet.

Zhao’s co-founder Simon went from “10× engineer” to managing multiple coding agents simultaneously, queuing tasks before lunch and letting them run. That’s the shift from individual contributor to orchestrator. The 700 agents at Notion are just an early signal of where headcount ratios are heading.

The Case for Decision Traces

Everyone agrees agents will change enterprise software. The debate is over who wins: incumbents that own the data, or startups that own the execution path.

Read more

Happy New Year, everyone!

In today’s edition, among other things:

• The State of Startups

• The State of Consumer AI

• The Next Mental State for Founders

• 2026 Predictions, Outlooks & Forecasts

Onwards!

Read more

In today’s edition, among other things:

• The Structure of the Fastest Startups

• Dreams about Robots

• AI Engineers and Equity

• Why humans are AI’s biggest bottleneck

Onwards!

The Structure of the Fastest Startups

For most of startup history, structure was something you worried about after product-market fit. You hired a few people, improvised roles, and promised yourself you’d “clean it up later.” That logic no longer holds. The fastest startups are designed to move quicker.

The fastest startups win because they treat structure as a throughput system for judgment, not a headcount plan.

NFX makes an argument for this shift in a simple idea: company design follows the constraints of the technology moment.

Founders often treat structure as fixed. In reality, org design is fluid. It bends to two forces: the technology moment you’re building in and the competitive pressure you face.

That’s a useful frame because it’s falsifiable. If your market rewards reliability and compliance, you will rebuild hierarchy. If your market rewards iteration, hierarchy becomes a tax.

Information travels instantly. Work gets executed instantly.

The essay points to “two pizza teams” and Atlassian “pods” as early answers to the same problem: move fast without bureaucratic drag. AI pushes this to the limit by making interdisciplinary work easier and by shrinking the minimum viable team.

AI has simply put this force into overdrive, with three main effects: 1. It has completed the flattening… 2. …teams can feasibly be smaller than ever before 3. This culture demands a different type of employee psychology.

The interesting claim here is not “teams are smaller.” Everyone repeats that. The claim is that coordination is no longer the reason you form a team. Innovation is.

And when innovation is the purpose, silos are self-harm.

It’s a new loop: diagnose → propose → execute.

A flat org is a way to compress the feedback loop until the market can’t keep up.

Flint argues early hiring logic flips: you don’t hire “Head of Sales” to build playbooks and manage process. AI will handle playbooks and repetitive tasks. You hire for results, taste, and insight.

Instead of designing roles around processes… you design roles around outcomes. AI handles the playbooks, and repetitive tasks. What you’re really hiring for is results, taste, and unique insight…

This lands because it matches what you see in breakout teams: the first non-founder hire is rarely a narrow specialist anymore. It’s a generalist with sharp edges. Someone who can write, ship, sell, and instrument. The role title becomes a promise, not a department.

You don’t hire for a role. The right person can be the role.

Your first job descriptions are culture documents. They tell the team what gets rewarded. The quote puts it cleanly:

“Broad roles are fine; vague outcomes are not.”

There’s another constraint NFX hints at: being in-person. They claim most early-stage teams slow down when they’re not together once markets shift. That’s probably directionally right, but it’s not universal. Deep research, commercialization, and distributed talent pools can beat co-location. The test is practical: do you lose days to misalignment, or minutes?

Two years ago, I would have disagreed. Today I agree in full.

What would change my view: clear evidence that remote-first teams consistently match in-person teams on cycle time during “unknown unknown” pivots.

So what’s the playbook?

• Write roles as outcome contracts. Define the one thing this person makes inevitable in 90 days. If it reads like a task list, you’re hiring a coordinator. If it reads like an impact statement, you’re hiring a builder.

• Design a team topology that minimizes handoffs. Keep the smallest unit able to diagnose, decide, and ship. If you need three people to run one experiment, you built a mini-bureaucracy.

• Screen for AI posture. NFX calls out “enthusiastic, informed humility.” In practice: watch how candidates use AI. Do they treat it like autocomplete, or like a junior analyst they can interrogate? Do they reject it on principle, or outsource thinking to it? Both are disqualifying.

The closing line of the essay is the right north star: in 1855, the job was coordination. In 2025, it’s creativity and speed.

The fastest startups look different because the constraint moved. The winners structure themselves around the new constraint before the market forces them to.

AI Engineers and Equity

For most of the last decade, equity followed a familiar script. Early engineers got 0.5–1%. Exceptional hires might stretch to 1.5%. That playbook is now broken. In AI-native companies, equity has stopped being a retention tool and started being a value signal.

Read more

In today’s edition, among other things:

• Things That Aren’t Doing the Thing

• The Tiny Teams Playbook

• Why You Can’t Focus at Work

• The Full-Stack Person

• How to Lead Remote Teams

• How to Set Goals for 2026

• How to Use AI Agents for Marketing

• How to Pick a Company Name

Onwards!

Things That Aren’t Doing the Thing

A reminder I’m reposting every now and then for everyone of us:

Preparing to do the thing isn’t doing the thing.

Scheduling time to do the thing isn’t doing the thing.

Making a to-do list for the thing isn’t doing the thing.

Telling people you’re going to do the thing isn’t doing the thing.

Messaging friends who may or may not be doing the thing isn’t doing the thing.

Writing a banger tweet about how you’re going to do the thing isn’t doing the thing.

Hating on yourself for not doing the thing isn’t doing the thing. Hating on other people who have done the thing isn’t doing the thing. Hating on the obstacles in the way of doing the thing isn’t doing the thing.

Fantasizing about all of the adoration you’ll receive once you do the thing isn’t doing the thing.

Reading about how to do the thing isn’t doing the thing. Reading about how other people did the thing isn’t doing the thing. Reading this essay isn’t doing the thing.

The only thing that is doing the thing is doing the thing.

The Tiny Teams Playbook

Seven teams. One hundred people total. Two hundred million dollars in ARR. That’s $2M in revenue per employee, a figure most venture-backed startups never approach. Something new is happening with AI: teams with fewer than 30 people are quietly shipping products used by tens of millions and generating nine-figure revenue. Turns out tiny teams work only when you treat headcount as a last resort and automate everything else.

It’s the architecture of those teams that makes speed possible.

In the Tiny Teams Playbook, seven companies with roughly 100 people in total generate about 200m ARR across products like Gamma, Bolt, Gumloop and Datalab.

Wang describes Tiny Teams as “the co-op multiplayer game… capable of more adaptability, resilience and ‘damage per second’.”

That metaphor is important. Tiny teams are tuned for output per unit of trust, not output per unit of time. Every extra person is another network edge to maintain. So they bias hard toward:

“Small (<15) crew of senior generalists: much fewer juniors”

You are not building a pyramid. You are drafting a raid party. Broad, experienced players who can own whole loops: talk to users, ship product, wire the data, fix the infra. Juniors do not vanish forever, but they arrive later, once the game is well defined.

That is, if you have the right kind of experience in building. You can build an organization that can assume agents and automation are the default and humans are the exception.

Here’s the advice from the companies that managed to do this:

• Hiring

  • Hire right or not at all: have to be excited about the candidate or it’s a no

  • Work Trials: paid projects for 4days-3months to be sure it’s a good fit

  • Product-Led Hiring: top customers who quit their jobs to join you

  • Top of market salaries: 95th+ percentile salaries

  • Small (<15) crew of senior generalists: much fewer juniors

• Culture & Value: keep a living culture deck and live it

  • Low ego, high trust: trust = speed, ownership

  • Independence, Grit & Resilience: ignore standard VC advice, persevere

  • Radical transparency and accountability: wall of work, show & tells

  • User focus: Work closely with users, celebrate them, delight in feedback

  • Camaraderie, speed: Have fun, do retreats, avoid burnout

• Operations

  • Almost no meetings: “deep focus” - building instead of talking about building

  • AI Chief of Staff: automate research, marketing etc w/ Gumloop or Lindy

  • AI Support: very well fleshed out at this point. e.g. see Parahelp and Railway

  • Let Fires Burn: in order to prioritize on the 10% critically important

  • Compound learning: Oleve phrases it “Don’t Learn It Twice” - build reusable templates and playbooks

  • In Person: either have an office, or VERY frequent AirBnB hack weeks

• Tech and Product

  • Simple, Boring Tech Stack: shell scripts over k8s, keep code modular.

  • Simple Product: start from UI wrapper over one API call to a LLM.

  • Feature Flags/Experimentation: one of Oleve’s core principles.

  • Benchmarks: create top tier internal evals for LLMs/harnesses. Market them.

The mechanism is simple: every recurring cognitive task is either eliminated, templatized, or given to an agent. The scarce resource in an AI native company is managerial bandwidth, not compute. Once you see it this way, it becomes obvious why these teams can stay tiny with huge surface area.

Why You Can’t Focus at Work

Microsoft’s 2025 Work Trend Index found that heavy collaborators face interruptions every two minutes. That’s 30 interruptions per hour. Run that through a basic productivity model and the output is zero: no 60-minute focus blocks, no deep work, no compound thinking. There are no productivity YouTube videos and books that can help you.

Our inability to focus isn’t a discipline problem, but a math problem, and the math says we’re operating in a system designed to prevent focused work.

Can Duruk’s simulation model reduces workplace productivity to three parameters: λ (interruptions per hour), Δ (minutes to recover focus after each interruption), and θ (minimum uninterrupted time required for meaningful work). The interaction between these variables explains why some days feel productive while others vanish into gray noise.

“Even a ‘quick two-minute question’ can cost you 15–20 minutes of recovery time.”

This asymmetry is the core mechanism. The interruption itself is trivial. The recovery penalty is catastrophic. At λ=2 (two interruptions per hour) and Δ=20 (twenty-minute recovery), Duruk’s simulations show workers getting roughly one 60-minute focus block per day. That’s from an 8-hour workday.

Seven hours of “working” produces one hour of actual work.

Read that again.

Seven hours of “working” produces one hour of actual work.

The relationship isn’t linear. Dropping from λ=2 to λ=1, just one fewer interruption per hour, increases the probability of getting three deep-work blocks from 14% to 70%. One interruption per hour is the difference between a sustainable system and a broken one.

Here’s a good day for comparison:

The simulations use λ=2 or λ=3 as “typical” scenarios. Academic research suggests these are best-case assumptions.

“González & Mark found workers switch activities every 3 minutes on average. Iqbal & Horvitz measured 7.5 email/IM alerts per hour, with 10–16 minutes needed to resume work after each.”

That’s from 2004 and 2007. Since then, we’ve added Slack, Teams, always-on video calls, and a culture that treats every message as urgent. The Microsoft data showing λ=30 for heavy collaborators isn’t an outlier. It’s the new normal for anyone in a coordination-heavy role.

The core shift is to treat λ, Δ, and θ as tunable variables.

Run the simulation at λ=15 with Δ=25 and the entire 100-day grid turns gray. No focus blocks of any length. The visualization looks broken because the reality is broken. We’ve normalized an environment where focus has been engineered out of the workday.

The González study found that nearly half of knowledge worker interruptions are self-inflicted. Checking email twice daily instead of continuously is the highest-leverage change most people can make. Train your teammates that your attention requires effort to access. This feels hard until you measure the alternative.

If your product or productivity strategy assumes deep thinking but your operating environment is calibrated for constant reaction, you have a math contradiction coming from not understanding what’s happening, not a culture problem.

First, attack λ. Second, reshape θ by decomposing work. Third, trim Δ. It’s math.

Really a brilliant analysis.

The Full-Stack Person

Contrary to popular LinkedIn posts, AI did not suddenly make everyone a 10x engineer. AI removed your excuse for not shipping.

The durable advantage now is becoming a full-stack person: someone who owns the problem end-to-end and uses AI to move faster across product, engineering, and customer reality.

Read more

In today’s edition, among other things:

• Pushing the Frontier is Expensive; Repeating it Isn’t

• What Books Actually Teach You About Being Wrong

• The Era of Physical AI

• The ElevenLabs Story

• The Enterprise Sales Playbook

• US Startups are Faster

• The Benefits of Bubbles

• The Ultimate PR Guide for Startups

Onwards!

Pushing the Frontier is Expensive; Repeating it Isn’t

The debate about “how smart” AI is misses the point. What matters is how quickly AI can take on longer, messier work and how fast the cost of that work collapses. If capability horizons keep doubling on a sub-year clock and prices fall ~10× per year once a frontier is reached, diffusion drives the economy.

Read more

In today’s edition, among other things:

• Extending Your Runway

• The Three-Player Game of Internet

• Five Levels of Work

• Identifying A Players

• Attention is Luxury

• The State of AI in Healthcare

• A B2B Founder’s Guide to Generating Demand From Scratch

• The State of Crypto 2025

Onwards!

Read more

In today’s edition, among other things:

• Lesson from Scaling

• The State of the Robotics Ecosystem

• LLMs are a Different Kind of Intelligence

• The State of Vibe Coding

• AI Software Development Stack

• Roadmap: European Resilience

Onwards!

Read more

1. For Leaders who want to implement AI in their organizations. Together with friends, we’re organizing the second edition of the AI_managers program, starting October 6th. So you don’t have much time to join. Join here.

2. For builders: Hackathon. Warsaw, October 18. Anthropic, ElevenLabs, Vercel, and more as partners. Top investors, including Credo Ventures, Dawn Capital, and more, including yours truly. Join here.

In today’s edition, among other things:

• The AI Coding Trap

• Steve Jobs on Hiring Great People

• Where Startup AI Dollars Actually Go

• The AI Healthcare Stack

• Shipping with Confidence

• Life Is Poker, Not Chess

Onwards!

The AI Coding Trap

We’re shipping more lines than ever — and still missing the thing customers actually feel: reliability. Demos are easy. Durable delivery is not. Code is cheap now; context, comprehension, and change-safety aren’t.

I’ve spent the last two years in code-gen AI, and code-gen Agents specifically and really loved this one:

The work isn’t “write code”; the work is “know what changed, prove it’s safe, and make it explainable.

LLMs collapse typing time. The bill shows up later, and it’s not the tokens—explaining what changed, stitching it in, and proving it didn’t break anything.

• Most software effort is thinking, not typing. With AI, the “thinking” shifts after the code appears, turning reviews and fixes into the dominant cost. That’s why “10× coding” often becomes ~10% delivery in practice. Velocity moved; the bottleneck didn’t vanish.

• The human experience matches the charts: fun parts go to the model; you inherit testing, deduping, docs, and deploys. The output is faster; the outcome isn’t.

This is the root of the difference between marketing copy that boasts… ‘10X faster’, and the marginal productivity gains in delivering working software… closer to 10%.

If you can’t say what changed, why it’s correct, and how you know, the speed is artificial.

There are two paths: ship fast and hope (vibe coding) or ship legibly and compound (AI-driven engineering).

• Treat current agents like ultra-fast juniors: breathtaking output, brittle understanding, and no real on-the-job learning. You must provide the learning system—context, constraints, and review.

• The curves diverge. Quick prototypes look magical; then complexity hits and the graph bricks. The compounding path is slower at first but survives complexity. Prototype speed is not product velocity.

Applications of sufficient simplicity can be delivered without the need for any human thinking at all… But you will hit a wall of complexity that AI is incapable of scaling alone.

The leverage is not “more code,” it’s making decisions legible across the lifecycle.

• Specification first. Narrow scope, enumerate invariants and edge cases before generation. Specs are the policy the agent can read.

• Documentation up front. Generate and review docs early; they become reusable guardrails and the memory agents don’t have.

• Modular design. Keep contexts small so any reviewer can reason about them quickly and test them thoroughly.

• Test-Driven Development. Ask the agent for tests before code; let tests guide implementation and catch regressions.

• House standards via context. Feed style and patterns; auto-lint and refactor on commit.

• Monitoring & introspection. Instrument logs and extract insights; make agent decisions observable and reviewable.

Value accrues in specs, tests, docs, and observability—assets that survive model swaps and team changes.

AI didn’t invent software engineering failure modes.

• The fix is cultural and procedural: enforce practices that minimize rework while growing shared understanding—code reviews, incremental delivery, modular design, TDD, pair programming, quality docs, CI. It’s classical engineering, adapted for agents.

• Known unknown: longer contexts and tool use might move the wall. But until agents can hold whole-system context and prove invariants, the wall moves; it doesn’t disappear. That’s why we built Shotgun.

What can you do?

Read more

In today's edition, among other things:

• Why Nobody Wants Your Product?

• The CTO’s Role in Early GTM Success

• The Post-AI Org Chart

• The Marketer of The Future

• How People Use AI?

• You're Not Busy, You're Emotionally Overwhelmed

• How Shopping Will Work In The Age of AI

Onwards!

Why Nobody Wants Your Product?

Customers don’t wake up thinking about your product. They wake up thinking about their own survival—revenue, costs, compliance, staff headaches. Your job isn’t to sell software; it’s to solve a problem urgent enough that someone diverts scarce time, money, and attention away from everything else.

This is why so many startups fail despite looking “logical” on paper. A better UI or a lower cost is not demand. The only thing that matters is whether you’re addressing a discrete, top-of-mind “job to be done” that your buyer already feels.

Three typical conditions to look for:

1. Urgency: The “hair-on-fire” problem. Without urgency, your ROI case doesn’t matter. One company built great retention software for retailers; customers agreed it added value, but it wasn’t even a top-five problem. Adoption stalled.

2. Repeatability: Can you find multiple customers in the same vertical with the same urgent need? Founder-led selling validates whether demand survives beyond friendly intros. Customer interviews don’t count until someone pays.

3. Scalability: Can you grow from some satisfied customers into many, with renewals and referrals lowering acquisition costs? Without this pull, growth is too expensive to sustain.

Urgency gets you in the door, repeatability keeps you busy, scalability makes you durable.

Founders often believe they’re in the business of building products. They’re not. They’re in the business of reducing pain. If the pain isn’t felt deeply and broadly, the product doesn’t matter.

Nobody wants your product. They want their problem to go away. The faster you accept that, the quicker you find fit.

Subscribe now

The CTO’s Role in Early GTM Success

Startups usually imagine GTM as a sales problem. In reality, the first unfair advantage often hides inside the codebase. The CTO’s early involvement in go-to-market isn’t a distraction from building — it’s how you build a repeatable growth engine from day one.

Unfortunately, most of the time it’s an underdelivered effort.

Technical founders bring rigor to what is often treated as “art.” Bowery Capital notes that CTOs can enrich datasets in Google Sheets with AI, identify the right ICPs, and filter by industry, size, and behaviors. Instead of guessing who to sell to, you’re running structured experiments.

• Clean ICPs mean fewer wasted conversations.

• Repeatable enrichment workflows mean pipeline is a system, not a one-off.

• Every test creates data that improves the next iteration.

If you’re doing it.

If yes, then technical speed turns into GTM agility: faster hypotheses, cleaner prospects, tighter loops. If you’re constantly waiting on deliverables, you move too slow.

It’s a myth that GTM is just “sales and marketing.” The best early motions blend product and market in real time.

• Engineers can automate enrichment, CRM updates, and outreach sequencing — saving reps from drowning in data entry.

• They design tests the way they debug code: controlled, measurable, repeatable.

• Most importantly, they translate features into value propositions prospects actually care about.

When they’re involved in GTM discussions, they can refine messaging and provide feedback to improve product-market fit.

That bridge between product and market often decides whether the first customers stick.

An often essential RevOps hire might take weeks to figure out how to link data sources and debug enrichment logic. A CTO can build the workflow in hours.

This turns them into the first true growth hacker: not the one sending cold emails, but the one architecting the system so the team can run faster with less friction.

In early GTM, leverage comes from systems, not scripts.

There are of course risks:

• Distraction risk: Pulling the CTO too deep into GTM can slow product velocity. The balance is making them the architect, not the SDR.

• Scaling risk: What works with a technical founder running Sheets may break once volume scales. You’ll need to hire for repeatability.

• Cultural risk: If engineers see GTM as “not their job,” the collaboration never sticks. Founders must set the tone.

The unknown: how far AI-native GTM systems can scale before hitting the limits of context, cost, and reliability.

Early on, the best advantage is your CTO turning technical intuition into a data-driven, repeatable revenue system. Why else ship anything?

Subscribe now

The Post-AI Org Chart

The org chart is the most invisible piece of software in a company. It encodes how decisions flow, what’s getting built, who gets promoted, and where bottlenecks form. But AI is rewriting that codebase. The AI-native organization won’t look like a pyramid — it will look more like a cylinder, where humans and agents share management layers in new ratios.

Read more

In today's edition, among other things:

• The Founder’s Invisible Tax: Cognitive Load

• How to Hire a Strong Founding Team

• How to Launch Your Startup Out of Stealth

• Developer Tooling for Software in the AI Era

• The AI Productivity P…

Read more

In today's edition, among other things:

• Fundraising Seasons

• Don’t Ship Your Pitch Deck

• Build vs Buy in the Age of AI

• Computer Use and Agentic Coworkers

• AI Creative Tools

• AI in Logistics

• Spec is The New Code

Onwards!

Fundraising Seasons

Fundraising has seasons, and your success rate depends as much on when you raise as on how you raise. Fundraising works best when you compress the process into the “hot” windows and plan ruthlessly in the off-season.

Venture capital doesn’t run on a continuous clock. It runs on pulses.

• Hot windows: January–March and September–November. Investors are back from holidays, focused, and primed to make decisions quickly.

• Cold windows: December and midsummer. Partners are away, ICs drag, deals stall. You can start a process then, but you’ll feel like swimming against the tide.

The same pitch delivered on September 10th versus December 20th lands in entirely different climates.

Momentum is oxygen in fundraising. Investors rarely act alone; they move when they sense others moving too. Here’s NFX:

“Momentum is the single most powerful force in fundraising. If you create it, your odds go way up.”

That’s why seasonal timing matters: everyone is active, everyone is watching. Staggered meetings in a dead zone stretch for weeks and kill urgency. A compressed two-week sprint in a hot season makes competition inevitable.

Seasonality is the amplifier, but the mechanics still matter:

• Fund math: Align your round size and ownership ask with target funds’ check sizes. A $5M raise at a $100M post won’t fit most Series A firms’ models, no matter the season

• Narrative clarity: As Sequoia advises, you have five minutes to hook an investor: what changed, what you do, and your fast facts. The rest is context.

• Handshake clarity: Y Combinator’s handshake protocol is simple: no confirmed offer until an investor explicitly affirms amount and terms in writing. It’s seasonality insurance against ghosting.

And yes, great companies raise year-round. But most companies don’t. For everyone else, seasonality tilts the odds.

• Raising in July can work if you compress and keep momentum while partners are out. Risk: drift. Reward: less noise.

• Raising across seasons is dangerous: if your round “hangs” into the next window, you look stale. Better to pause, sharpen metrics, and relaunch fresh.

The unknown: whether agent-powered fundraising (AI-native sourcing, automated IC prep) will flatten these cycles.

To recap:

• Now (end of off-season): Prepare. Deck, metrics, warm intros. Back-solve from September or January.

• Next (on-season): Compress. Run parallel meetings to get 3–5 firms term-sheet-ready within 10–14 days.

• Later (long-term): Build fund-math awareness. Track what ownership targets your likely leads need, and design your ask to fit.

Good luck.

Don’t Ship Your Pitch Deck

Staying on the fundraising side a bit more. Founders often confuse the story they use to raise capital with the one they need to win customers. Your pitch deck sells investors on a vision; your website must sell customers on solving a problem today.

Emily Kramer and First Round has seen this mistake across hundreds of startups:

A great pitch deck makes for an ineffective website. It also makes terrible outbound copy. You can’t copy-paste that deck to your homepage because your prospects likely aren’t VCs.

A deck is aspirational—“system of record,” “$100B market,” “world-class team.” Customers don’t buy aspirations; they buy relief from pain. They want to know what your product does right now, why it’s better than the alternative, and how to get started.

The temptation is to mimic Stripe or Figma, aiming to sound like a scaled company. That’s a trap.

Your early advantage is focus. You can solve a very specific problem for a very specific audience better than anyone else. Lean into that. Don’t work against yourself by trying to appear like you solve everything for everyone.

This means:

• Pick a narrow wedge problem.

• Speak in your customer’s language, not VC jargon.

• Avoid category-creation theatrics until you actually have traction.

Early marketing isn’t about vision; it’s about precision.

The smart move is to treat each pitch slide as raw material—not finished messaging. For example:

• Problem Slide: For VCs → “Radiologist-read images often misinterpreted.” For customers → “Reduce diagnostic errors”.

• Solution Slide: For VCs → “Compliance and security management platform, starting with SOC 2.” For customers → “SOC 2 audit platform”.

• Vision Slide: Investors want the $10B TAM. Customers want “We manage the customs process for everything you ship”.

• Market Slide: Swap “remote-first companies globally” with “Built for early-stage remote startup teams”.

My biggest learning is this: assume zero context. Spell it out.

If your first 10 customers can’t tell what you actually do from your site, no investor deck in the world will save you.

Build vs Buy in the Age of AI

For as long as I’m in tech “build vs buy” has been a binary debate in tech. Either you build in-house (flexible but costly) or buy off-the-shelf (fast but constrained). That trade-off doesn’t go away becouse of AI but it’s different now.

Read more

In today's edition, among other things:

• AI Commerce

• The Future of Software Business Models

• How to Articulate Your Thoughts More Clearly

• Agents vs Reality

• Al for HR

• Start Earlier

• State of Cybersecurity

Onwards!

Read more

In today's edition, among other things:

• How Social Media Shortens Your Life

• AI Defensibility: From Speed to Staying Power

• Enduring Companies in the Age of AI

• The Paradox of Expertise

• Open-Weights vs. Open-Source Models

• AI Is Polytheistic, Not Monotheistic

• Incompetency Framework: The Truth About Skills-Based Hiring

Onwards!

How Social Media Shortens Your Life

We rarely think about time as a consumable good. Yet it’s perishable, non-renewable, and unlike money you can’t earn it back. Pretty obvious. This makes it strange how casually we give it away. Social media doesn’t just take your time; it changes your sense of what time is.

The theft is very subtle. You don’t notice a few seconds disappearing in a checkout line, so why notice a few minutes in an app? But the architecture is deliberate. As Sean Parker put it when describing Facebook’s original design philosophy:

How do we consume as much of your time and conscious attention as possible?

This is why a half-hour on Instagram feels like a sip of water: you don’t remember taking it. Gurwinder calls this the 30-minute ick factor, “the sickly feeling you get after looking up from your phone and realizing that 30 minutes have passed without you realizing it.” He adds:

A social media feed is like the Lethe, the mythological river of forgetfulness, in whose waters lost souls sought absolution—and received it in the form of oblivion.

There’s another trick: what you remember isn’t the time you spent, but the moments that broke the pattern. Narratives and novelty stick; the rest is blur:

The way you perceive the passage of time depends less on the actual amount of time that’s passed, and more on the amount of your experience that your brain remembers… when you can’t remember much of what you did, your brain concludes that not much time passed, and so it seems like your life was shorter.

Scroll long enough and you erase the peaks and valleys that mark time in memory. Life starts to feel shorter because fewer days stand out. And when time feels short, the brain often misreads it as time well spent.

This is one of social media’s most insidious tricks: the more of your life it takes, the more it makes you feel like it’s giving you.

The result is a kind of mental inflation: each unit of time buys you less experience. Worse, the damage doesn’t end when you close the app. Constant interruptions fragment your ability to pay attention to anything else, corroding the quality of your off-screen hours. As he writes:

If you spend enough of your attention on that which you’ll quickly forget, you’ll end up quickly forgetting your life.

The fix isn’t simply “less screen time.” It’s adding back texture: right-angle turns in your routine, conversations with real stakes, activities that will be remembered next year.

You can’t store time. But you can store moments. And moments, unlike hours, are what actually add up to a life.

Source: Gurwinder.

AI Defensibility: From Speed to Staying Power

One of the most interesting discussions I had this year was with Aydin Senkut, founder of Felicis. Very inspiring and smart. Here’s from his Linkedin post:

Abundant capital doesn’t get evenly distributed, it gets concentrated. Ruthlessly !!

Escape velocity attracts capital gravity. It no longer matters whether a company is public or private. If you're growing fast enough – 100x user growth, $0→$100M ARR in a year, redefining categories – capital will find you. From every direction. VCs, crossover funds, strategics, sovereigns. The laws of capitalism guarantee it.

The total pool of companies that reach that threshold is shrinking. The ones that breakaway are shipping in beast mode. A year of progress now takes months. What took months now takes days. And days will soon collapse into hours.

So, it’s ruthless, fast-growing, sharp-elbow market out there. it’s not enough to scale fast — you need to build a fortress, not just sprint through the gates. How?

Read more

In today's edition, among other things:

• The Autonomous GTM

• The Handshake Deal Protocol

• How to Present to Investors

• How to Reindustrialize a Country?

• Work Never Gets Finished When You Are Drowning

• The Future of Software Development

Onwards!

Read more

In today's edition, among other things:

• The SEO for AI

• The Autonomous Enterprise

• Winning in the US as a European Startup

• Writing Code is Not the Bottleneck

• The Quitting Point

• Why Voice Will Be the Fundamental Interface for Tech

• The AI Agent Market Map

• You Have 5 Seconds

• The New SaaS Org Chart

Onwards!

The SEO for AI

Although some companies are still lying to themselves, the fight is no longer over search‐engine results— that’s over. What large language models remember and repeat - this is the most crucial aspect to consider going forward.

For 25 years, ranking meant deciphering Google’s algorithm. The durable asset is no longer link equity - it is model memory. Whoever lives in the model’s “memory” owns the discovery and will soon see both ads and checkout in all major LLM clients, starting with ChatGPT.

Traditional search parsed links; generative engines parse language.

• Queries have stretched from four words to roughly twenty‑three as users converse instead of typing. Sessions stretch to six minutes because follow‑ups are free and immediate.

• LLMs look up sources, reason across them, and answer. A brand’s goal shifts from winning a single query to being a reliable ingredient in the model’s synthesis pipeline.

• Formatting matters: bullet lists, clear headings, and semantic density make content easier for models to chunk and quote. Counter‑intuitively, over‑optimized keyword salad actually decreases the odds of mention because it lacks narrative coherence.

Click‑through rate (CTR) ruled dashboards. AIO/GEO (AI Optimization, Generative Engine Optimization - people are still not aligned on how to call it) substitutes them with

• Reference Rate – the percentage of model answers that cite or paraphrase your content unprompted. Canada Goose, for instance, discovered that when ChatGPT voluntarily named the brand, unaided awareness spiked in human panels a week later.

• Attribution Depth – whether the model quotes you superficially (“warm parka”) or leans on proprietary insight (“triple‑stacked down baffles tested at ‑30 °C”). Deeper attributions correlate with higher conversion on subsequent clicks.

• Outbound Share – Semrush clickstream data shows ChatGPT referrals jumping 300 % in five months, topping 30 000 unique domains per day. Sites that appear early in that outbound tree enjoy first‑mover advantage as the model reinforces its own patterns.

In short, you are optimizing for mindshare inside silicon, not shelf space on a results page.

Here are the most popular tools for GEO:

Read more

This is a summer edition of the newsletter and I’m currently off on short holidays. Every time I take some time off I return to some important and often evergreen articles, videos and books.

Weekly analysis will return in the next week edition.

In today's edition, among other things:

• How to Success and Thrive in Career You Love

• How to Do Great …

Read more