TLDR Founders 2026-08-03

QM is a multiplayer agent harness for companies. It remembers enough for people to work together without treating the whole company as one person with one permission set. The agent recognized personal, channel, group, team, and organization scopes, allowing people to collaborate through the same agent without becoming one undifferentiated user. QM is open-source and available for Slack and the web.

Model pricing can now go stale before a startup ships. OpenAI cut Luna's API price 80% only three weeks after launch and dropped Terra's price 20%. Anthropic released Opus 5 at half the price of Fable 5, while Google priced Gemini 3.6 Flash below Kimi K3 per task. Cheaper models are already forcing frontier labs to move before most enterprises have even started switching. If model spend matters to your margins, keep switching costs low.

Service is what the customer experiences when trying to get something done. For people to trust you, you need to make their journey easy, respect their time, take ownership, ensure that advisable details inspire confidence, be competent and decent when something goes wrong, and build a system that learns. Trust takes time to earn but can be lost surprisingly quickly. It is one of the most valuable things a service provider can possess.

Forward-deployed engineers brought proximity, speed, and ownership of the outcome, but that isn't enough. Engineers are still often too far away from the key decision-makers who are critical for decisions and culture. Companies can't just rapidly hire and replace the entire C-suite in a month - they need outside help. Forward-deployed executives can help close the gap by bringing both technical and leadership experience to help businesses navigate the chaos and complexity of AI adoption.

QM is an agent designed for startups. Each employee gets their own isolated workspace and can work independently without affecting each other. They can also collaborate with the agent in channels, group messages, and projects. Skills are scope-owned and shareable by grant. QM can be used to search internal knowledge, build internal apps, work on existing repositories, and track projects in a shared channel with updates and follow-ups.

OpenAI's new campaign type, Agent, allows marketers to send ChatGPT searchers and users to a Business Agent conversation. The agents will be created through a combination of scraped data from the business website and instructions from advertisers. They will have access to product feeds and MCP tools, and can capture leads through custom forms. They can also launch ad campaigns that point to the agent rather than a website URL.

Most of what founders currently call a moat is a lease. Routing, distribution, marketplace liquidity, and the app-layer feature set all lose value as the models get better. What appreciates is proprietary data, expert decision-making, and a pipeline that turns those decisions into the next training run.

Over 380,000 founders, operators, and investors read TLDR Founders to stay on top of the latest trends, tactics, and news in the startup ecosystem. If you are a founder or investor and want to help curate it, send your LinkedIn or resume to founders@tldr.tech!

Startup options can turn into a bill the day an employee leaves. The usual 90-day window asks someone to pay the strike price and possibly taxes before they know whether the stock will ever be worth anything. That made more sense when startups exited in four to seven years, not today's 10 to 20 years. Backplanes gives people their tenure minus one year to exercise, while Craftwork gives them a full decade. If you use equity to recruit great people, make sure they can actually keep it.

Manifest ran an LLM router for four months across 7,000 users, then decided to shut it down.

WhatsApp is testing a feature that automatically moves messages from large businesses to an 'Offers & Updates' folder after a set number of hours.

Some quick tips on what investors look for before wiring money.

Drive can come from pain, but it can also come from curiosity, high standards, or a family that never learned to coast.

Patrick Collison talks about dropping out of MIT twice, why founders should ask what happens if they succeed, and what Stripe's own data says about the best time to start a company.