Claude · Anthropic

Split work between three founders and Claude to compete with giants

Paxa Labs, a three-person Thai text-to-speech startup, says it competes with much larger AI companies by letting Claude lead coding, documentation and testing while the founders keep Thai-language research for themselves.

What they did

The team split work three ways. Research that needs deep Thai-language context is human-led. Coding, documentation and tests are Claude-led, including rewriting existing components in Rust and checking that outputs still match the original. Marketing, business conversations and content are drafted by Claude and then reworked by the founders.

How it went

The founders say their Thai voice model now beats larger providers on latency, speed, naturalness and price, especially on Thai place names, dates and numbers.

Worth knowing

The comparison is the team's own claim from a meetup talk, reported in an attendee's recap with no benchmark numbers. They note Claude sometimes pushes back and that soft-skill work still needs people.

Try it with Claude

We are a <team size> person team building <product>. Here is our current backlog: <backlog>. Sort every item into three groups: work you should lead (code, docs, tests), work we should lead (research needing <domain> expertise), and work you draft and we finish (marketing, partner emails). For the first group, propose an order and the checks you will run to prove nothing broke.

Swap the parts in angle brackets for your own details. This prompt is adapted from the case, not copied from the original.

What this label means: User story

Someone using it at work posted about it. Any numbers are theirs, so we write "the poster says".

Summarised from the original source; the link has the full story. The people and companies here are not Wize customers.

Want this done for your team?

Wize sets up and runs agents for your team on the platforms we manage, and a person approves anything that goes out.

See the platforms we run