ByteDance has ample computing power and AI talent, but one analysis argues that competition between internal teams may be getting in the way of building large models. The question is whether the company’s “horse-racing” approach can work for research that depends on teams sharing resources and coordinating closely.
ByteDance AI Watch analysis
What happened
A 27 October essay published by 36Kr argues that ByteDance’s internal competition model, successful in app development, is a poor fit for large-model research. It points to teams competing for computing resources and says short assessment cycles can reward visible results over work such as data cleaning and infrastructure optimisation. Read the essay at 36Kr.
The essay, written by Tech Business and published with its authorisation, says ByteDance has since brought parts of the Seed organisation closer together and ended quarterly OKRs and half-year assessments for the Seed Edge project. Those are the essay’s claims; they are not independently established by the supplied evidence. The piece also recounts allegations of deliberate interference with model training, which we do not repeat as established fact.
Why it matters
Large-model development is not just a matter of adding GPUs or running more experiments. If teams cannot share infrastructure, data and fixes, a company may spend heavily while making it harder to coordinate the work that makes those resources useful.
The essay’s argument is broader than ByteDance: structures built for fast product competition may not suit long, tightly coordinated research. But it is analysis, not proof that organisational design explains ByteDance’s progress or that its rivals have solved the same problem.
Our read
The useful question here is not whether internal competition is always bad. It is whether teams are competing in ways that leave shared infrastructure, data quality and sustained research nobody’s job. That is a more consequential test than counting GPUs or teams. The essay gives readers a case to interrogate, not a verdict to rubber-stamp.
What to watch
- Whether ByteDance confirms the reported Seed and Seed Edge organisational changes.
- Whether researchers describe improvements to shared computing, data and infrastructure work.
- Whether the essay’s claims about internal competition are supported by further reporting.
Discussion spark: For large-model research, should a company prioritise competition between teams or central coordination of shared resources?
Sources and evidence
- ByteDance Large Models: Trapped in Internal Horse-Racing Mechanism – Largest Computing Power Yet Slowest Progress – eu.36kr.com (27 October 2026, 00:52 UTC)
- AI levels field for regional start-ups, say experts – Gulf Times (27 October 2026, 00:52 UTC)
not affiliated with or endorsed by ByteDance AI