About
What the dimsum is, which regions and topics it covers, how every claim is verified before publication, and how corrections ship — in public, dated.
the dimsum is a twice-weekly publication on AI engineering intelligence from labs across China, India, the US, and Japan.
The frontier conversation tends to process these labs through press releases and leaderboard screenshots. We go the other way: model cards, technical reports, arXiv papers, repository contents, rate cards, and license files — read directly, every issue, and synthesised for engineers who act on the details.
What we cover
- Model architecture — attention and memory-system design, MoE routing, multi-token prediction, long context.
- Inference and serving — kernels, frameworks, speculative decoding, pricing surfaces that matter for agent loops.
- Open weights — licenses read from the artifact, checkpoints inspected, serving-stack support checked.
- Independent evidence — benchmarks cross-checked, vendor numbers always labelled as vendor numbers.
How we work
Every issue follows the same pipeline: research → verification → editorial synthesis → human review → publication. Every checkable fact traces to a ledger row with its source and evidence tier before it enters a draft — see Methodology for the full description.
Issues publish twice weekly, on Wednesdays and Sundays; each issue covers all four regions. The feed is available as RSS.
Corrections
We do not silently edit material factual errors. Corrections ship in a dated Corrections section at the bottom of the affected issue, with the primary source each correction was verified against.
Frequently asked questions
What is the dimsum?
the dimsum is a twice-weekly publication tracking AI engineering intelligence from labs across China, India, the US, and Japan.
Which topics does the dimsum cover?
Model architecture — attention and memory-system design, MoE routing, long context; inference and serving — kernels, frameworks, speculative decoding, agent-loop pricing; open weights — licenses read from the artifact, checkpoints inspected; and independent evidence — benchmarks cross-checked, vendor numbers labelled as vendor numbers.
How are claims verified?
Every checkable fact gets a row in a claim ledger recording its source, evidence tier, and decision cost before it enters a draft. Nothing is published without human editorial review. The full pipeline is on the Methodology page.
When does the dimsum publish?
Twice weekly, on Wednesdays and Sundays. Each issue covers all four regions and is permanent at its dated URL. The feed is available as RSS.
How are corrections handled?
Material factual errors are never silently edited. Corrections ship in a dated Corrections section at the bottom of the affected issue, with the primary source each correction was verified against.