Robert Ta breaks down Cursor's experiment rebuilding SQLite six times with different AI models, revealing costs from $1,339 to over $20,000 for identical work. Using Opus 4.8 and Composer 2.5 as a planner-worker swarm cut costs to one-twentieth of single-model runs. The real bottleneck in future engineering won't be code generation but writing precise specifications.
Cursor rebuilt SQLite 6 ways and the costs range from $1,339 to over $10,000 on the exact same task. And for a lot of people out there, we've talked about this before and people show up in the comments and they're like, well, it was trained on SQLite. It knows about SQLite. It's like, yeah, that's not really the point of these benchmarks. Like a benchmark is a benchmark. It's not perfect. But what these benchmarks do, and I'm glad Cursor did this research, is because we're builders trying different metas, they actually experimented and showed the cost efficiencies or lack thereof of this particular meta of let's Let's just let a frontier intelligence run away and rebuild SQLite. You see Fable 5 at $20,000 in cost, GPT-5 at about $10,000, Opus 4.8 at $5,000. So the question becomes, can we get the same result using a different meta? What Cursor indeed found is you could do this, like Opus 4.8 and Composer 2.5 together as planner and worker. You can get it down to 18th to 20th of the cost. And I thought this was really interesting is the agent swarm concept. They talk about specs as prompts. Each jump in AI capability has raised the level of abstraction at which an engineer can work. Autocomplete let engineers work one line of code at a time. Early models raised that to a block of code, and agents raised it to a file or a feature. With swarms, the unit of work becomes the spec. For that to work, the swarm has to actually follow the spec. In all of traditional software development, it always comes down to how good are the requirements, how well understood is the specification, and if you have hundreds of engineers working on it, are they actually following the spec? So Cursor says we gave the swarm 835 pages of And it came back with a database. What was scarce in this experiment and what we expect to be scarce in software engineering going forward is the right description of intent. Seen this way, the swarm starts to resemble a compiler. A compiler translates source code down to machine code through a series of intermediate steps. Planners parse a goal into task trees, then lower it step by step into executable work. The difference is that a compiler preserves meaning at every step, while the swarm is probabilistic at every one. They actually have the Swarm output here. You can go take a look at their SQLite implementation from this 830+ pages of prose.
