Anthropic run rate: $67B
Anthropic's revenue run rate grew from ~$60B to $67B even after DeepSeek's shock debut, showing cheap Chinese models have not yet collapsed American lab economics.
Snapshot · This Week in Tech (Audio)
TWiT 1093: California Sober - Kimi K3, Qwen3.8, & China's Open-Weight AI Gambit
Explore episode
Jul 20, 2026
Where this was said
Dean Ball's Thread: Open-Weight AI Is 'AI Communism' and a Dystopian Hellscape
At 51:12 · chapter starts 47:50
Leo Laporte walks through Dean Ball's influential tweet thread, now that Ball has become OpenAI's head of Strategic Futures. Ball's argument has four points: Kimi K3 is genuinely impressive and cannot be explained away by distillation; China allowing its release is strategically surprising; open-weight models deter AI capex by making private investment in frontier models irrational; and an open-weight dominant world is essentially AI communism — state-provided digital infrastructure rather than a market product. Harper Reed takes sharp issue with Ball's framing, calling the 'China is just too stupid to know what it's doing' angle a 'fatal flaw' in most US analyses of China. Leo reads Ball's final warning: at some point these models will be capable enough that a non-living, self-replicating agent escaping from a Chinese lab will be a real concern. Harper responds that US labs haven't been blameless about training data copyright either, and that the real question is how America competes in a game it's not currently playing.
The GLM Revelation: Chinese AI Was 6 Months Behind Claude Code — Now It's Free
Harper Reed used Claude Code back in early 2025 when it was still rough. Today's GLM is at least as good as that version of Claude — and it's effectively free. The only question is how much time separates frontier from open-weight models. If the gap is 6 months and the open-weight version is free, why pay?
DeepSeek V4 at $0.12/M tokens
DeepSeek V4 costs ~$0.12 per million tokens, vs $10+ for Claude Fable, illustrating the dramatic price gap between Chinese open-weight and American closed frontier models.
Open-Weight Models Already Good Enough for Most People — The Frontier Gap Is Closing Fast
For everything except hard-core coding, open-weight models are already good enough. Harper Reed's Toyota Corolla analogy nails it: a Corolla is a completely rational car. It does the thing. You might prefer something else, but 'better' doesn't mean 'worth paying 50x more for.'
Similar snapshots
500K impressions from one tweet
A single post tapping into the AI coding debate drove close to 500,000 impressions, making it the founder's best-performing piece of content.
100x easier to join existing trends
The founder argues it is 100 times easier to bring your ideas to where attention is already focused than to create attention from scratch.
Build-in-public content rarely goes viral
Most founders building in public never go viral because they never join the bigger conversation already happening in their space.
3 years of content creation before product launch
The speaker built his audience over 3 years of consistent content creation before launching any product.
5 minutes a day to tweet
Tweeting consistently took the speaker only 5 minutes a day, making audience-building accessible to anyone.
Audience = prerequisite for product revenue
Having an existing audience was cited as the primary reason the speaker was able to make significant money from a product launch.
Content formats: YouTube + Twitter
The speaker recommended creating YouTube videos and tweeting as the two core content formats for building an audience.
Elon Musk estimated SpaceX had 5–10% survival odds
After SpaceX's third rocket failure, Elon Musk estimated survival odds at only 5–10%, yet stated no failure probability would have caused him to walk away — a textbook example of religious-stage commitment.
$14K/month in 6 months
Sam built Algrow from zero to $14,000 in monthly revenue within just six months of shipping his first MVP.
