TL;DR
Matt Wolfe covers 18 AI news stories from the week, including DeepSeek V4's near-state-of-the-art performance at a fraction of competitor pricing, the Elon Musk vs. Sam Altman trial going poorly for Musk, Anthropic's billing controversy over harness detection, and Google signing a Pentagon AI deal despite employee protests.
Key points
1
DeepSeek V4 is open-source with 1M token context, priced at $1.74/M input tokens vs GPT-5.5 at $5/M and Claude Opus 4.7 at $5/M, while performing near state-of-the-art on benchmarks
2
The Musk vs. Altman trial began this week; Musk performed poorly on the stand, refusing yes/no answers and arguing with attorneys, with one Verge reporter saying she had never been more sympathetic to Sam Altman
3
Anthropic was caught charging users extra or blocking them when code contained strings like 'Hermes' or 'OpenClaw', affecting Claude Max subscribers; they eventually issued refunds after the story went viral with over 1M views
4
Google signed a deal with the Pentagon allowing AI use for any lawful government purpose, contradicting a 2014 promise made to DeepMind founders and drawing a 600-employee protest letter to Sundar Pichai
5
Mayo Clinic developed an AI model that can detect pancreatic cancer on routine CT scans up to 3 years before clinical diagnosis, identifying disease before tumors are visible
Actionable insights
→
Enterprises spending heavily on closed API models like GPT-5.5 or Claude Opus should evaluate open-weight models like DeepSeek V4 or Mistral Medium 3.5 for significant cost savings and privacy benefits
→
Developers using Claude Code should audit their repos and commit messages for strings like 'Hermes' or 'OpenClaw' to avoid unexpected billing routing or service blocks from Anthropic
→
OpenAI models are now available on AWS following the restructured Microsoft partnership removing exclusivity, meaning multi-cloud AI deployments are now more viable
Notable quotes
“After about 5 hours into Elon Musk's testimony, I typed the following sentence. I have never been more sympathetic to Sam Altman in my life.”
“Most of the time what we get is like, hey, making Excel spreadsheets is a little bit easier now. We need more stories like this to come out, like actually detecting cancers earlier and curing cancers and solving real actual problems.”
“There is a certain class of bugs that suggests the thing you are trying to do is a bad idea. Worth reflecting on that.”
Worth watching?
⏭️
Worth watching the full video?
All 18 stories are fully covered here with the key facts, pricing figures, and drama details intact — skip the video unless you want to hear Matt's commentary and see the live demos.
Saved you some time? The creator still deserves a like.
Want this for your own channels?
Add the channels you follow. Every new video summarised and in your inbox the moment it drops. From £4/month.
Try it free