TLDR Founders 2026-06-24
Selling inference β‘, first deals π€, company building companies πΒ
Most founders understand their finances in spreadsheets β and that's a problem. (Sponsor)
Your bank has the data. But to actually use it β check your cash position, reconcile transactions, understand your burn β you export it, paste it somewhere else, and do the work outside your account.
Mercury Command closes that loop. It's AI built directly into Mercury* that surfaces insights and takes action from your live account data. Ask what you need to know, then act on it instantly β follow up on an outstanding invoice, set a limit on a card, categorize a transaction β all in the same conversation. You approve every step. Command executes.
Try Mercury Command β
*Mercury is a fintech company, not an FDIC-insured bank. Banking services provided through Choice Financial Group and Column N.A., Members FDIC.
So You Want to Sell Inference (3 minute read)
AI companies selling inference face a zero-margin, cost-plus pricing dilemma. They can achieve better margins through value-based pricing, charging based on outcomes rather than raw inference costs. Optimizing inference costs using model routing, caching, and distillation can also enhance margins and offer a competitive edge.
Companies That Build Companies (6 minute read)
Polsia, a startup with AI agents instead of employees, claims $10M in annual revenue, illustrating the rise of companies building companies using AI. Both Polsia and YC-backed Thomas focus on AI-driven business models promising to innovate startup creation. While most ventures may fail, the few successful ones could prove lucrative, resembling Shopify's model where only a small percentage of ventures succeed significantly.
7 Mistakes Founders Make When Chasing Their First Deals (8 minute read)
Founders often delay selling for product perfection, underprice their offerings, and rely too much on a single customer. They should identify decision-makers, listen more than they speak, and adapt their pitch continuously. Early-stage technical founders must personally handle sales to fine-tune their approach before hiring dedicated salespeople.
How to Steal a Customer Base (9 minute read)
The fastest way to take a competitor's customers is to leave them nothing to walk away from. Cursor didn't ask developers to switch off VS Code, it forked VS Code, so every extension, shortcut, and bit of muscle memory came along for free, then rebuilt the engine around AI underneath. Microsoft can't simply copy that back, because rebuilding VS Code that deeply would put its whole extension ecosystem at risk, so it won't, and that refusal is the counter-positioning that holds the door open.
What you need to know about Lambda MicroVMs (7 minute read)
Lambda MicroVMs provide VM-level isolation with fast startup and stateful sessions. They are useful for coding assistants that need to run AI-generated code safely, interactive AI notebooks, agent sandboxes, and vulnerability scanners. The VM automatically suspends when idle and resumes when traffic arrives. Vertical scaling within a MicroVM is automated.
Introducing Claude Tag (5 minute read)
Slack users can now tag Claude in chats on selected channels. The bot can connect to tools and data to complete requested tasks. It builds context by remembering relevant information from the channels it's in, and it can plan out tasks to complete in the future. Claude Tag is now available in beta for Claude Enterprise and Team customers.
Collaborative Posts Are Coming to LinkedIn β What They Are and How to Make Them Work (6 minute read)
LinkedIn will soon implement a feature that allows two or more people to share one post together, with every collaborator listed at the top. The feature is currently in beta with creators and brands. It will roll out over the coming months.
AI Agent Hype Meets Reality: The Product Doesn't Work (And Churn Is Coming) (5 minute read)
Instead of building a superagent that does everything, make the focus as narrow as possible. Make sure your product has been extensively tested and that it works in real-world conditions. Companies that promise everything but produce AI slop will lose customers quickly. This churn makes growth impossible to maintain.
Every Moat Becomes Moot (8 minute read)
A good reframe of what a moat actually is. It's not a static wall like a patent or a network effect, it's an engine that only keeps working if you stay willing to do the uncomfortable, repetitive thing rivals won't. Amazon is the example, the thin margins that look like weakness are the wall, because fat margins are really just a fund for your competitors' R&D. The moat that lasts isn't any single advantage, it's being able to build the next one faster than the current one erodes.
Get our free, 5-minute newsletter read by 200,000 startup founders, entrepreneurs, and CEOs
Join 380,000 readers for
one daily email