Austin walks through an election day full of AI news. Anthropic warns its AI may soon speed up its own development, AI super PAC ads barely mention AI, and Apple's record quarter comes with a profit warning driven by AI-fueled memory costs. Also covered: the small nonprofit feeding chatbots their voting facts, an MIT experiment where one AI edited another, and Cardi B telling college students not to ask ChatGPT for answers.
Watch from 1:55
- What happened
- In an updated stock filing released Monday morning, Anthropic warned that future models may soon cross a safety line the company drew itself, tied to AI accelerating its own development. No current model has crossed that line. The filing says Anthropic has missed big jumps in its own progress until after the fact. It also shows $17 billion in sales over three months, more than the entire first half of the year, and a $567 million operating loss mostly from compute costs.
- The claim
- Anthropic says AI may soon be doing much of the model-training work that used to require human researchers, but that no current model has crossed its safety threshold.
- Austin's take
- Austin frames this as recursive self improvement (RSI) and calls it an exciting step forward. The catch is verification: humans checking AI-built work is slow, which can undercut the speed gains. He argues that if RSI is already here, political efforts to slow AI (like fighting data centers) may barely matter. He also notes a stock filing is partly a sales pitch.
- Why it matters
- If AI starts doing most of the work of building the next AI, progress could outrun the people responsible for checking it, and the company itself admits it has missed big jumps before.
- What to do
- Watch how labs describe their safety thresholds and human oversight in future filings and releases. Treat financial disclosures as both information and marketing.
- Not confirmed yet
- Austin mentions rumors that other labs, including Google, have already achieved recursive self improvement, based on Google's fast jump to a top model. These are unconfirmed rumors.
Watch from 9:29
- What happened
- Bellingcat studied 336 ads from four AI-funded super PACs. Only ads for two candidates mentioned AI, and none mentioned data centers. Most focused on the economy, immigration, and Trump. Bellingcat says a pro-AI network took $125 million from OpenAI's president, his wife, and a VC firm, while Anthropic gave $40 million to a rival pro-regulation group. Both sides adjusted their Trump messaging depending on the audience. An advocacy poll says three quarters of voters find AI money a turn off.
- Austin's take
- Austin thinks AI should not be a political issue and calls the industry money in elections outrageous, while agreeing data center siting and environmental rules need real attention. He doubts elections will slow AI much and sees politicians on both sides jumping on the issue without strong convictions.
- Why it matters
- The AI industry is spending heavily to shape elections without telling voters the ads are about AI.
- What to do
- When you see a political ad, check who paid for it, not just what it talks about.
- Not confirmed yet
- The show gives $125 million for the pro-AI network in the main segment but $150 million in a promo read at the end. The voter poll comes from an advocacy group.
Watch from 15:03
- What happened
- Apple posted record sales of $114.5 billion on Monday afternoon, then warned holiday profit margins will shrink. The stock fell as much as 4% before recovering some losses. The new CEO called memory prices unprecedented, as AI data centers soak up the world's memory chips. Apple also pays more to run Siri AI in the cloud. Apple says it has already raised some prices and won't say whether more hikes are coming.
- The claim
- Apple describes memory prices as unprecedented and declined to say whether further price increases are planned.
- Austin's take
- Austin says businesses will keep paying big markups for AI chips, but consumers will push back on much pricier iPhones. Long term he thinks Apple is very well positioned to win consumer AI by running models locally on devices, which is better for privacy, and calls Apple probably a buy right now.
- Why it matters
- The AI boom is now raising the cost of the phone in your pocket, not just data center budgets.
- What to do
- If you're planning a device upgrade, expect prices to stay high or rise. Austin's stock view is opinion, not financial advice.
- Not confirmed yet
- The new CEO's name is pronounced inconsistently in the episode (Turnis and Kernis), so check the spelling elsewhere.
Watch from 21:59
- What happened
- NBC News reports that Democracy Works, a small nonprofit in Brooklyn, supplies election data used by Google, OpenAI, and Anthropic. It pulls voting rules from state and local election offices, and its own site lists deals with those three companies. The group says it is struggling for money and isn't sure it will survive until 2028.
- The claim
- Democracy Works says it sources voting rules from state and local election offices and is unsure of its survival to 2028.
- Austin's take
- Austin notes chatbots are great with historical facts but weaker on live information, since they need web search. He wonders if there will be a market for real-time models that are trained continuously.
- Why it matters
- Millions of voters ask chatbots about polling places and deadlines, and those answers may hinge on one underfunded group.
- What to do
- For voting details, confirm with your state or local election office rather than trusting a chatbot alone.
Watch from 25:40
- What happened
- An MIT team used OpenAI's GPT-6 Astra to directly edit a small Qwen model (about 800 million parameters) without training. The edit touched far less than 1% of the model and its math score rose about six times. The team says the edit stopped the model from looping on answers. A Hugging Face engineer rebuilt and reran the test and found the gain came mostly from stopping endless loops. The method failed to help a different small model.
- Austin's take
- Austin calls it a phenomenal breakthrough and a step toward AI improving itself. He sees it as a way to get more out of small local models and spread that power to individuals who could use a big model to tune a small one for their own needs.
- Why it matters
- It hints that large models could make small, cheap, local models more capable, though the result is narrow so far.
- What to do
- Treat this as an early, isolated result. Watch for whether the approach works on other models before expecting practical tools.
- Not confirmed yet
- Independent replication found the gain was mostly from fixing looping, and the technique did not help a second model.
Watch from 31:10
- What happened
- Cardi B performed at Johns Hopkins' Hoptoberfest after stepping in last minute for Daniel Caesar. She told students to pass their semester and added, "don't ask ChatGPT for the answers." In February she joked about freeing older relatives from Facebook AI. Some colleges now give students AI tools for class.
- Austin's take
- Austin likes the message: learn the work hands on first, then automate. He sees having AI write your essay as cheating, but using it for research with sources, or as a tutor that teaches you how to solve problems, as smart rather than cheating.
- Why it matters
- Schools are split on AI, and the line between help and cheating is blurry for students.
- What to do
- If you're a student, check your class's AI policy and use AI to learn the material, not to hand in its output as your own.
- Not confirmed yet
- Austin says he thinks New York State or New York City banned AI for students but wasn't sure which.