Thoughts About Scaling Law
Scaling, but not only of parameters. Every model release now ends with the same question: how many parameters? It isn't a question that can be answered on its own. Parameter count is only meaningful alongside three others — how much data you have, where you intend to spend your compute, and who will run the model, under what conditions.
The field learned this the hard way. Kaplan et al. (2020) fit an exponent that told everyone to grow parameters faster than data — roughly 2.7:1 — and the industry complied: GPT-3, Gopher, MT-NLG. Hoffmann et al. (2022) redid the experiment across four hundred models and found the compute-optimal split is closer to 20 tokens per parameter, and that with sufficient compute the two should grow at the same rate rather than drifting apart. The error in the earlier fit compounded with every order of magnitude of compute, which is why the largest models of that generation were the most misallocated. The trillion-parameter round was, in retrospect, a detour the whole field took together and then reversed.
Chinchilla wasn't the end either. It optimized training compute for models that would be trained once and evaluated. Today a model is called billions of times a day and inference dominates lifetime cost. Put inference into the objective and the optimum moves toward smaller models trained far longer — deliberate over-training, which is what Llama-2-7B and Gemma-2-9B were doing at roughly 290 and 889 tokens per parameter.
Sparsity moved the target again. In a MoE model two quantities have to be kept apart: total parameters govern roughly how much the model can hold — knowledge, facts, the long tail — while activated parameters and effective depth govern roughly how far it can think, how many steps of a causal chain it can carry before it comes apart. A dense 20:1 ratio does not transfer. And the ratio isn't a single number at all: Roberts et al. (2025) find the optimal tokens-per-parameter is task-dependent, with memorization favoring more parameters and reasoning favoring more data. Follow-up work on MoE observes that at fixed TPP, pushing total parameters higher actually degrades reasoning, while activating more experts reliably helps it.
This matters for what we are building toward. Finding a vulnerability is not a retrieval problem. It doesn't come from having memorized more CVEs; it comes from carrying a twenty-step chain of inference to the end without losing the thread. That capability does not live in total parameter count.
Which brings us to this release. Total parameters appear to matter up to a threshold — enough to hold the world — after which additional capability comes from scaling elsewhere: effective depth per forward pass, and above all post-training. GLM-5.3 is our controlled experiment on that claim. Same base, same architecture, same total and activated parameters as GLM-5.2. One month of scaling long-horizon environments and RL. The gains are not marginal. Well, scaling has more than one dial. We turned the post-training one this time because it had the most slack left in it — not because the others are finished. Base model size, pretraining data, compute spent per forward pass: all of them are still on the table, and we will come back to each. What this experiment taught us is that the dials do not have to be turned together, and that the one worth turning next is rarely the one that was worth turning last. We are not done scaling. Next time, maybe mid-training, pre-training, and even more.
Claude Sonnet 5 is crazy efficient for web deisgn.
❤️🔥Just recorded a 18-min tutorial on how to build award-winning websites with Claude Code + Sonnet 5!
Announcing mattpocock/skills v1
- Achieved a 63% reduction in token cost for skill descriptions
- Split skills into model-invocable and user-invocable skills, adding /codebase-design, /domain-modeling, and /grilling
- (UPDATED) /writing-great-skills - rewritten from the ground up, encoding my skill-writing best practices
- (UPDATED) /diagnose -> /diagnosing-bugs - now model-invocable, awesome for fixing hard bugs
- (NEW) /ask-matt: a router skill that teaches you how all the engineering skills work together
Google 把内部工程师的代码审查(Code Review)规范公开啦
这几乎是目前业界最顶级的标准
很多程序员只会写代码,但不知道怎么审代码,可以看看 Google 是怎么做的
1.双向指南:不仅教审查者怎么挑毛病,还教作者怎么写出容易通过的代码
2.术语科普:解释了 Google 内部常用的 LGTM(看起来不错)和 CL(变更列表)到底意味着什么
3.实战价值:这套规范不是理论,而是 Google 每一位工程师都在用的实际操作准则
如果你想提升团队的代码质量,或者想知道顶级大厂的开发门槛,这份文档必读!
https://t.co/OdaozRkMYn