Shawn Wang (swyx) - founder of smol.ai, Latent Space, AI Engineer, DX.tips
Shawn “swyx” Wang explains how he deliberately created the AI Engineer movement, why AI changes the product requirements for DevTools, and what founders misunderstand about DevRel, angel investing and working on the right things.
Swyx is the founder of smol.ai, cohost of Latent Space and the organizer behind the AI Engineer Summit and World’s Fair. He also invests in developer infrastructure and previously led developer experience work at Netlify, AWS and Temporal.
This return conversation ranges from building a professional category to personal agency, LLM-native infrastructure and the post-ZIRP reckoning in DevRel. The thread connecting it all is intentionality: choosing a game worth winning, then designing the systems that make progress toward it more likely.
The Q&A below is experimental — extracted from the episode transcript with AI. If you spot a mistake, let us know at jack@scalingdevtools.com.
What is the ambition for the AI Engineer conferences?
The first Summit gathered about 500 people and the next World’s Fair about 2,000. Swyx models the larger event on practitioner gatherings such as KubeCon, AWS Summits, GDC and Strange Loop: a definitive annual place where an industry changes jobs, finds collaborators and launches companies. The Summit can remain focused while the World’s Fair grows toward the 5,000–30,000-person scale that would make AI engineering an established professional community.
Was the AI Engineer category an intentional launch?
Entirely. A February 2023 hackathon convinced Swyx that software engineers building AI products needed a category distinct from ML researchers and prompt engineers. He considered names including LLM Engineer and Cognitive Engineer, chose the simpler term, then coordinated the essay, conference launch and podcast positioning before publishing. What later looked inevitable began as one of the most deliberately sequenced launches he had attempted.
What does it mean to work with principles rather than plans?
Plans change at first contact with reality, but planning is still valuable preparation. Principles provide the recurring direction. Swyx’s deeper interest is intentionality or high agency: after establishing enough personal and career security to take risks, ask whether each opportunity belongs on your plate at all. Winning on purpose may mean joining the right team rather than founding a company; the point is to choose meaningful work instead of inheriting somebody else’s ladder.
What are the two accountability questions?
First: what is the most important thing you should be working on? Second: why have you not done it yet? Swyx imagines a private, non-judgmental support group organized around those questions. Most productivity systems optimize execution after the goal has been chosen, while too little time goes into strategy. Failure usually comes from choosing the wrong work or leaving a concrete psychological, financial or practical blocker unresolved.
Could an AI be the CEO of one person?
Swyx’s “small CEO” idea inverts the usual hierarchy. A human board sets values and goals, a machine maintains state and accountability, and the human becomes the executor. Managers and founders are paid to remember ownership, status and deadlines, but some people are naturally better at reacting than maintaining that state. AI could keep important but non-urgent work visible—including “invisible deadlines” whose opportunity window becomes obvious only after it has closed.
Are developers going to become obsolete?
Swyx argues that AI engineers would be among the last roles replaced because they perform the final automation work for every other AI-enabled field. Systems engineering also demands precision, planning and performance that language models do not automatically supply. Automation can expand an industry before reducing a role, as ATMs initially accompanied growth in bank branches. AI DevTools have an additional advantage: the builder and user can be the same person, producing an unusually tight feedback loop.
What makes a DevTool native to LLM users?
A tool that works well for humans may be a strong base—Supabase gives models a well-known interface to Postgres—but machine customers change the operating envelope. Agents may need hundreds of parallel sandboxes, much lower latency and error messages written as corrective prompts rather than human prose. Even API keys may need agent-specific permissions and human review for irreversible actions. The opportunity is often a familiar infrastructure primitive redesigned for orders-of-magnitude more machine use.
Why become a DevTools angel investor?
Not for quick returns: successful positions may remain illiquid for a decade. Swyx began with his own money, then became a scout so a venture firm could handle the legal, tax and administrative machinery. A public voice and industry network create deal flow; the portfolio becomes evidence for a possible future investing career. More immediately, advising founders on launches and DevRel exposes him to company-building problems before he encounters them himself, turning the relationship into mutual preparation.
Is DevRel dead?
The title contracted, but the job of making people know and care about a product did not. The zero-interest-rate version—expensive travel, weak attribution and activity disconnected from revenue—faced a necessary reckoning. Some companies can reach a few major customers directly and do not need a dedicated role; others still underinvest in telling the market what they build. Swyx places DevRel primarily at the top of the funnel, while acknowledging that teams retreat toward safer, more attributable customer education and success work.
Who should an early-stage company hire for DevRel?
The founder should first do enough of the work to understand what succeeds and create a repeatable playbook. Swyx’s ideal hire is an engineer from a lighthouse customer who has personally persuaded a company to pay for the product. That person understands the technical experience and the commercial stakes. A large free-tier audience alone is not the same capability. For smaller startups, coding ability also gives the hire credibility and a path into engineering if the role changes.
Links & references
- The Rise of the AI Engineer ↗
- Latent Space ↗
- smol.ai ↗
- swyx on X ↗
- swyx’s website ↗
- DevTools Angels ↗
- DX.tips ↗
- Is DevRel Dead? ↗
- AI Engineer World’s Fair ↗
This episode is brought to you by WorkOS.