Selling inference ⚡, first deals 🤝, company building companies 🔁
This issue highlights AI-powered banking insights via Mercury Command, strategies for profitable inference pricing, and the rise of “company-building” startups. It also covers common founder missteps in early sales, tactics for poaching competitors’ users, Lambda MicroVM use cases, new AI features like Claude Tag, LinkedIn’s collaborative posts, agent product pitfalls, and why moats demand ongoing effort.
- Charging for raw inference compute caps margins; pricing per user action or business outcome is what actually makes AI products profitable
- "Company-building" startups like Polsia (claiming $10M ARR) are using AI agents instead of employees to launch multiple companies at once, betting on a few Shopify-style breakout winners
- Cursor beat Microsoft by forking VS Code itself—keeping all extensions/keybinds intact while adding AI—something Microsoft can't copy without breaking its own ecosystem
- Real moats aren't patents but continuous, unglamorous work competitors won't bother doing, requiring constant reinvention as the edge erodes
Mercury Command plugs AI directly into your bank account—no more exporting statements to spreadsheets. From one chat interface you can check cash, categorize transactions, cap card spending, even chase unpaid invoices. It sits inside Mercury (a fintech platform backed by Choice Financial Group and Column N.A., Members FDIC) and executes only with your OK.
AI startups selling inference are stuck on razor-thin margins if they bill by raw compute. Shifting to outcome-based pricing—charging per user action or business metric—can unlock real profit. Teams also cut costs by routing requests to smaller models, caching frequent queries and distilling big ones into leaner versions. On the flip side, Polsia and YC-backed Thomas are building “companies that build companies.” They spin up new ventures using AI agents instead of hiring staff, with Polsia already claiming $10 M in annual revenue. It’s a high-risk play: most won’t make it, but a handful could mirror Shopify’s 10x winners.
Founders chasing first deals trip over three big mistakes: waiting for a perfect product, underpricing to win foot-in-door sales, and relying on one “anchor” customer. The fix? Get in front of decision-makers fast, listen twice as much as you talk and tweak your pitch with every call. Meanwhile, Cursor stole Microsoft’s VS Code users by forking the editor—preserving every extension and keybind while slipping in an AI engine underneath. Microsoft can’t retrofit without breaking its ecosystem.
On the tech ops side, Lambda MicroVMs offer lightweight VMs that spin up in milliseconds, stay stateful and pause when idle. They’re ideal for safe code execution in AI assistants, sandboxed agents and vulnerability scans. Slack just added Claude Tag, a beta feature for Enterprise and Team plans that lets you tag Anthropic’s Claude in channel chats; it remembers context, hooks into your tools and schedules follow-ups. LinkedIn’s next beta will let multiple authors share one post, listing all collaborators at the top. But buyer beware: AI agents that promise every feature rarely work in the real world. Products that overpromise and underdeliver will see churn spike. And don’t mistake a moat for a patent—real defense comes from doing the endless, low-margin work rivals won’t, then reinventing that edge before it fades.