Search the archive

ORIGINAL THINKERS

The standard in venture used to be having a thesis and a network.

Check out the Persona Startup Program today→

CLOUDED JUDGEMENT 5.1.26 - THE DEATH OF PER-SEAT PRICING?

The entire modern SaaS industry was built on the per-seat licensing model. However, Microsoft says that model is dying. In the company's new model, the seat remains, but it becomes a packaging mechanism for prepaid consumption. Beyond a certain level, users pay per token, per agent action, or per outcome.

AFTER “AI-FIRST” COMES “AI-ONLY”

'AI-first' asks whether AI can do the work of the human pilot. 'AI-only' means extensive workflows in which no human sits inside the operating loop. Humans still set the goals, values, constraints, and conditions for escalation, but execution runs end-to-end on machines. This shifts humans from participating in workflows to being a steward of the system.

EVERYONE WANTS TO BE AI-PILLED. MOST COMPANIES ARE STILL LEVEL 1

Miura-Ko adapts the AV autonomy framework to AI adoption, sorting companies into six levels based on what AI can see, do, who can extend it, and how the org chart has changed. Most "AI-forward" companies still sit at L1 personal productivity or L2 functional silos, while L3 requires agents acting across CRMs, code, and tickets through MCP and shared skills authored by non-engineers. L4 introduces a compounding operating system where finance staff and AEs ship production internal tools and policy-driven agents fix issues with human review only at merge, while the still-hypothetical L5 has the system noticing problems, acting within delegated authority, and updating shared memory without a human initiating the loop.

THE SMALL INTERVIEW MISTAKE THAT QUIETLY COSTS FOUNDERS BIG MEDIA

Most founders spend their time preparing for interviews when they should be preparing for the next invitation instead. They try to use their time to sell their company when instead they should intentionally answer questions asked and connect their points to the conversation without forcing it. Audiences can tell when a founder is just there to sell something. They can also tell when someone is there to actually teach, inform, or have a real conversation.

WHAT'S WORKING RIGHT NOW IN AI SEARCH

Buyers are showing up to your homepage with the comparison already done. They ask an assistant to shortlist vendors, explain tradeoffs, and rule out obvious mismatches before clicking anything. Your positioning got rebuilt from whatever the model could find: blog posts, support docs, comparison pages, YouTube clips, third-party reviews. One company doubled lead quality by editing pages LLMs already cited, not by chasing more citations. Another had so much old content that AI answers described the product wrong. Messaging drift used to be a brand problem. Now it shows up in shortlists.

THE LENS PREMIUM

AI made competent content cheap. What's still scarce is a clear way of seeing the market that customers start repeating back to you. A blog, podcast, founder account, or event doesn't compound because it's "useful." Useful is everywhere. It compounds when it gives buyers a frame they didn't have, a way to name what they were already feeling. The Stripe and Notion versions of this look like content strategy, but they're really positioning. Name the customer's half-formed belief better than they can, and they'll start using your words to explain the problem.

START YOUR FREE 14-DAY TRIAL TODAY HERE.

RELEASES

Linear Releases integrates with CI/CD tools to precisely track the deployment environment, version, and status of every issue. It gives team members and agents full deployment context so they can see what's live to customers, not just what's been merged. Releases will automatically issue status updates when their associated code changes land in production. It supports both continuous deployment and scheduled release methodologies.

ADD CUSTOM AGENTS TO PRIVATE SLACK CHANNELS

Notion's Custom Agents can now read and reply in private Slack channels. They can only see the private channels they are invited to. The Slack API Connector allows users to find quick answers in Notion using information from Slack conversations. The feature is only available on the Business and Enterprise plans.

IF AI IS SO GREAT, WHY ISN'T IT WORKING?

Models are already smart enough. Enterprises shouldn't be holding out for AGI. The companies that build the operational layer underneath the models, as opposed to waiting another five years for frontier AI to improve, will win the next decade.

AI GTM REFACTOR: 5/3/2026

AI models are now good and cheap enough. There are great tools that can give any company a competitive edge. It's easier than ever to build. Take an AI-pilled customer-facing engineer and give them a Vercel account and unlimited token spend. You'll be surprised what they can ship in a weekend.

BECOME A CURATOR FOR TLDR FOUNDERS (3 HRS/WEEK)

TLDR is looking for someone who has founded a VC-backed startup to help write TLDR Founders for hundreds of thousands of subscribers. Our curators have been invited to major tech events, scouted for Tier 1 VCs, and get early access to unreleased TLDR products. Learn more [18].

WHEN YOUR VC LEAVES, YOU ARE FUNDRAISING AGAIN

A founder's responsibility to act in the best interest of their company and its shareholders doesn't change just because a partner has left.

WHY DOES NO ONE IN AI HAVE SUPPORT?

Almost no non-enterprise AI apps have any real form of support.

MOST COMPANIES AREN'T ANYWHERE NEAR READY FOR AI

IT'S ALWAYS ABOUT VELOCITY

Speed only tells you how fast you are moving, whereas velocity tells you how fast you are moving and where you are going.

PAYING TO WIN: THE RISE OF VEBLEN SERVICES

Veblen services are professional services where price signals credibility and assurance. At the top end of several industries, high fees are a signal of quality. Hiring the most expensive firm becomes a form of insurance. In this environment, lowering fees raises doubts, and discounting looks more like a weakness.

THE INFERENCE ECONOMY: TOKEN USE

Token demand is going up. Everyone should be thinking about how to manage token usage as models will get more expensive as they get better. Businesses need to consider whether they are using the right-sized model for their project and whether reasoning is really required. They need to be flexible with providers and only use fine-tuning/post-training if they have enough data, expertise, and resources.

THE MOST FUN I'VE HAD BUILDING APPS: GPT-5.5 + GPT-IMAGE-2

Paste a screenshot of an app into Codex. Codex builds it, then opens a browser and compares the running app against the original screenshot. In the Artemis II example, Codex pulled real NASA mission data, noticed the screenshot's Earth-to-Moon ratio was wrong for the actual mission, and interpolated a flight path that preserved the visual without lying about the physics. The system treated the visual as design intent, not gospel. The bottleneck stops being "can you make the first version" and starts being "can you tell what's off."

THE NEW UNDERUTILIZED SALES TRICK YOU CAN STEAL

Add an "Ask AI" button to your landing page. The button opens ChatGPT, Claude, or Perplexity with a pre-filled prompt about your product. The user clicks, the AI explains, and the explanation feels less like marketing because it didn't come from you. A landing page is obviously self-interested. An AI assistant feels like a second opinion. The execution matters. If the prompt says "tell me why this product is amazing," it reads as fake. The stronger version asks what a real buyer would ask, including where the product might not fit. The balanced answer is the persuasion.

X LAUNCHES THE MOST AMBITIOUS ADVERTISING PLATFORM REBUILD IN ITS 20-YEAR HISTORY

X has completely rebuilt its advertising platform from the ground up. It started rolling out the new platform last month. The new Ads Manager is built around the core pillars of Simplicity, Unmatched Control, and Superior AI-Powered Performance. The upgrade introduces modern retrieval and ranking systems powered by state-of-the-art AI. It enables more precise, relevant, and dynamic ad delivery that aligns with what's happening on X in real-time.

LINK CLI

The Link CLI allows agents to complete purchases on users' behalf without ever storing their real card details. Agents receive secure, one-time-use payment credentials from a Link wallet, so payment credentials are never exposed. Users have the option to make sure the tool requests approval for every transaction.

INTRODUCING ADS CLI: A COMMAND-LINE INTERFACE FOR META ADS AND COMMERCE

The Meta ads CLI is a command-line tool that allows developers and AI agents to manage Meta ad campaigns directly from the command line without writing custom code. It packages the Meta Marketing API into one tool with predictable commands that both developers and AI agents can use reliably. The tool supports multiple output formats and can integrate seamlessly into existing workflows.

A COSTUME CALLED CONVICTION

Most VCs call themselves conviction investors. What they usually mean is they'll move quickly when a founder fits an obvious pattern: AI company, elite school, prior exit, hot market, clean memo. Real conviction is expensive. The founder is strange, the market is unproven, and the memo is ugly. The piece names this the "legibility tax." Every layer of institutional capital strips away weirdness so the deal can be defended internally. If you're building something non-obvious, you need one person willing to look foolish before the room agrees.

PRE-SEED IS NOT CHEAP SEED INVESTING

Besvinick rebuts Lucas Bagno's "Great Bifurcation" thesis that low-priced rounds signal adverse selection, arguing pre-seed at $5M-$15M post operates on a different risk profile than $40M-$100M seed rounds and rewards backing founders before consensus forms.

MAKE FRIENDS WITH YOUR MARKET

Most teams pay for the booth, then waste the room. They sit behind a table, check their phones, and wait for obvious leads to walk up. Meanwhile, the openings are everywhere. Someone slowing down for half a second. Holding eye contact. Touching a brochure. The move isn't a pitch. It's an easy invitation. The piece also makes the case for talking to neighboring booths, not just buyers. One 20-minute conversation with the booth next door led to an intro into a customer segment the company hadn't considered.

7 ATTRIBUTES OF AI-NATIVE STARTUPS

AI-native startups should leverage data as their primary competitive advantage, establishing data flywheels to outperform competitors.

EVERY BUBBLE LOOKS DIFFERENT

Every cycle invents a new name for the same fantasy: software that makes money while you sleep.

YOU CAN MAKE SOMETHING GOOD

Make things you are personally curious about and put them in front of people fast.

Skeptical? START YOUR FREE 14-DAY TRIAL

2026: THE YEAR OF CHURN

"Does it integrate with Salesforce?" was the question buyers asked for a decade. Now, they want to know if agents can drive your product, if your APIs are clean, and if there's an MCP connector. Buyers asking those questions are buyers running the math on whether to keep you. The math gets worse when a vendor they already pay adds a 60% version of what you do, because AI fills enough of the rest that cutting your $80K contract stops feeling like a tradeoff. Annual deals have hidden how often this is already happening.

SOFTWARE IS EATING THE WORLD (BUT ACTUALLY THIS TIME)

AI agents now execute the actual work as code loops rather than just mediating interfaces, converting human jobs into tool calls and verification steps. Agentic tasks burn orders of magnitude more tokens than chat because a single Claude Code bugfix can consume around 900k tokens, where almost all goes to context replay and tool output rather than visible code. METR data shows autonomous task horizons doubling every 131 days, jumping from 4 minutes on GPT-4 to roughly 12 hours on Claude Opus 4.6 and pushing OpenAI past 15B tokens per minute alongside Google's 50x YoY token growth.

A NEW TOKEN RULE FOR ENGINEERING LEADERSHIP

Chainguard now expects its engineering managers to be at the 50th percentile of token usage among their direct reports. Managers who use too few tokens lack the context needed to manage and move results forward. Managers who use too many tokens need to focus on increasing token usage amongst their teams. Right around the middle means that both managers and their teams understand what their tools can do.

THE FOUNDER'S GUIDE TO BLOGGING

Polished prose used to take effort. Now, any founder can produce a clean feature announcement in ten minutes, but most of them read like the same announcement. Swap your company name for a competitor's and read your last post again. If it still makes sense, you wrote a press release. Don't hand the model a feature spec and ask for a polished post. Hand it the customer ticket, the Slack argument, and the voicemail where you explain why this even matters.

are limited. Request to attend today

FOUNDER MARKETING SKILLS

ScaleBrick's Founder Marketing Skills offers four key marketing skills for founders: social search audit, growth strategy, keyword research, and competitor audit. It supports a wide range of agents without requiring accounts or API keys, and has driven significant growth for companies like a YC-backed healthcare app and AIFlyer.

REASONS BLOCKS

Catch failures mid-run, cut tokens, and compound intelligence across every agent you deploy.

SURETHING.IO

Run and coordinate AI agents that execute tasks and report results with shared context.

THE BRAVE NEW WORLD OF AI MARKETS

It is naive to assume that the Total Addressable Market (TAM) for artificial intelligence will simply equal the current size of the global human labor market. This post proposes a new framework for sizing AI markets based on the classic economic equation of Price multiplied by Quantity (TAM = Human Labor TAM × Price Compression Factor × Volume Expansion Factor).

DIRECTION OVER SPEED

For most of software's history, you could fake productivity by shipping a lot. AI broke that cover. Prototypes are cheap, specs write themselves, and shipping looks the same whether you're building the right thing or the wrong one. Smaller teams of four or five, with the PM closer to the work, especially on AI features where "is this good" can't be answered by a metric. The model can build the thing for you. It still can't tell you why anyone should care.

SHIPPED ISN'T SOLVED

A customer asked for a feature on a call. The team built it, shipped it fast, and moved on. Nobody really used it. Speed only helps if you understand the problem before you start building. Look at your last five shipped features and count how many customers were still using each one a month later.

BECOME A CURATOR FOR TLDR FOUNDERS (3 HRS/WEEK)

TLDR is looking for someone who has founded a VC-backed startup to help write TLDR Founders for hundreds of thousands of subscribers. Our curators have been invited to major tech events, scouted for Tier 1 VCs, and get early access to unreleased TLDR products. Learn more [18].

STOP ASKING YOUR DATA QUESTIONS

The things you most need to know are usually the things nobody thought to ask about.