Pretrained

Richard and Pierce respond to the Times podcast about the scarcity of junior engineering jobs. They talk through the academic difference between Computer Science vs. Engineering, AI as a new engineering primitive, talent arbitrage through intern programs, and more.

https://www.nytimes.com/2025/09/29/podcasts/the-daily/big-tech-told-kids-to-code-the-jobs-didnt-follow.html

OpenAI & NVIDIA’s 10GW partnership, GDPVal as a new human curated benchmark dataset, Gemini Robotics-ER 1.5, and Apple's distillation of AlphaFold.

Additional reading:
https://nvidianews.nvidia.com/news/openai-and-nvidia-announce-strategic-partnership-to-deploy-10gw-of-nvidia-systems
https://openai.com/index/gdpval/
https://deepmind.google/discover/blog/gemini-robotics-15-brings-ai-agents-into-the-physical-world/
https://arxiv.org/pdf/2509.18480

Pierce and Richard recap Anthropic's Economic Index. Differences between country use of AI, autonomy versus augmentation, and the real business use cases that Anthropic is seeing so far.

Further reading:
https://www.anthropic.com/research/anthropic-economic-index-september-2025-report

The official Claude Code post-mortum, Deepseek R1 published in Nature, Meta unveils their smart glasses with built‑in display, the new apple pro, copyright law in the age of AI, and much more.

Further reading:
https://www.anthropic.com/engineering/a-postmortem-of-three-recent-issues
https://www.nature.com/articles/s41586-025-09422-z#MOESM1
https://www.theverge.com/tech/779566/meta-ray-ban-display-hands-on-smart-glasses-price-battery-specs
https://www.reuters.com/legal/government/anthropics-15-billion-copyright-settlement-faces-judges-scrutiny-2025-09-09/

The Anthropic economic index report, a bug in claude's inference pipeline, OpenAI releases a flavor of GPT-5 just for coding, Microsoft's new inhouse LLM, and what really happens when you turn temperature to 0

Pierce, Richard, and Will join for the first in-person interview on Pretrained. For our video episode, check out:
https://youtu.be/CInTOIgz-pA

They cover:
- Will's history growing up in the UK
- Getting an MBA and deciding what company to start
- Trust building versus activating content
- Building a personal brand for engineers and researchers
- Entrepreneurship in Europe vs the US
- & Much more

Kimi's serving architecture, mooncake to offload GPU memory to other chipsets, the ubiquity of vllm, and the growing standard LLM stack

Kimi K2 and Moonshot AI's history, avoiding loss spikes during training, the muon optimizer, and data parallelism

Gemini’s new image model, OpenAI is investing more in protein generation, Cohere’s SOTA generation model, and Anthropic working with DOE on nuclear security

Further reading:
https://blog.google/products/gemini/updated-image-editing-model/
https://openai.com/index/accelerating-life-sciences-research-with-retro-biosciences/
https://cohere.com/blog/command-a-translate
https://red.anthropic.com/2025/nuclear-safeguards/
https://www.picolm.io/demo-paper

Rich and Pierce speculate wildly (well, kind of wildly) on the internal architecture of Genie 3. They go into the history of variational auto encoders and diffusion models, 3D modeling with AI as an alternative to video game designers, a recap of the official Genie 1 paper, and possible applications of world models to the real world.

Further reading:
https://deepmind.google/discover/blog/genie-3-a-new-frontier-for-world-models/
https://arxiv.org/pdf/2402.15391