Austin closes out the week with six AI stories: ChatGPT ads that can target you by email or phone number, Amazon's $75 listening wristband, a Federal Reserve survey suggesting AI is slowing hiring rather than causing layoffs, Meta's free Muse Spark Open model, Accenture capping staff Claude spending at $29 a month, and new voices and accents in ChatGPT's voice mode. Ron was off getting coffee, so it's a solo show with plenty of audience questions.
Watch from 2:25
- What happened
- Advertisers can upload email addresses or phone numbers to aim ChatGPT ads at specific people on the free tiers. 404 Media tested it by buying ads warning that humans read chats. Opting out of personalized ads does not stop this targeting, and ads can still follow what you type in the current chat. Uploaded lists must include more than 25,000 people, and one tracker has logged over 43,000 ChatGPT ads. OpenAI would not say whether it scrambles raw emails and phone numbers the way Meta does.
- The claim
- OpenAI says advertisers never see which people matched and that uploaded lists are usually deleted within a day.
- Austin's take
- Austin sees ads as a natural move for a company under revenue pressure, pointing to reported OpenAI spending of up to $700 billion through 2030, over a billion weekly users, and Anthropic reportedly posting record, profitable numbers this week. He asked viewers whether ads will feel normal or ruin the experience.
- Why it matters
- Your contact info can now follow you into a chat app, much like Facebook ads, and the usual opt-out does not fully turn it off.
- What to do
- If you use free ChatGPT, assume ads may be targeted using your contact info and what you type in the current chat. Be careful what you share, and consider whether a paid tier is worth it for you.
- Not confirmed yet
- OpenAI's financial position and spending figures were described as reported, not confirmed. It is unclear whether OpenAI hashes or scrambles uploaded contact data.
Watch from 19:18
- What happened
- Amazon launched the Bee Frontier, a $75 wristband that listens to your conversations and turns them into summaries and to-do lists. It lasts up to a week per charge and has a light that shows when it is listening. A $100 phone-back version, the Bee Eclipse, is coming later, and a Bee Pro subscription costs $12 a month. It plugs into Alexa and Amazon shopping. The people around you have not agreed to be transcribed.
- The claim
- Amazon says no audio is ever stored and only you can read your transcripts.
- Austin's take
- Austin sees real value in giving AI agents context from your offline life, like a conversation about booking a babysitter that your agent could then act on. But he thinks Amazon is late to a crowded field of similar devices and admits he is skeptical of big tech promises not to mine your data. He also noted the irony of the launch coming the same week as the story of a spy planting a listening bug in OpenAI's offices.
- Why it matters
- Always-on wearables record more than just the wearer, which raises consent questions for everyone nearby, and they tie your daily life directly into Amazon's shopping ecosystem.
- What to do
- If you try one, tell the people around you when it is on, and read Amazon's privacy terms rather than relying on the marketing line.
Watch from 31:32
- What happened
- The Fed's latest survey of businesses found that 5 of 12 regional districts reported firms cutting staff through attrition, delaying hiring, or expecting fewer jobs after investing in AI. In Dallas, AI agents ran 180 hours of job interviews in two days. Minneapolis staffing firms said AI is not causing layoffs but is dampening new job creation. A Dallas accounting contact said big firms are delaying entry-level contracts partly because AI makes them more efficient.
- Austin's take
- Austin thinks this is the start of a longer trend. It is hitting repetitive, well-defined entry-level work first, like voice agents handling accounts receivable calls, and he expects accounting and bookkeeping to be next once people trust the tools. He hasn't seen it affect anyone he knows, and he argues that getting really good at AI tools is the best way to stay ahead.
- Why it matters
- If you are job hunting, especially for entry-level roles, openings may shrink quietly before anyone actually gets laid off.
- What to do
- If you are early in your career, build real proficiency with AI tools in your field. If you manage a team, think about how to retrain and reposition people rather than just leaving roles unfilled.
- Not confirmed yet
- The findings are anecdotes from business contacts, not a job count, and overall employment was roughly flat.
Watch from 15:02
- What happened
- Meta released Muse Spark Open, a free, downloadable model built for coding and long tasks. Apple tuned its software so the model runs on its desktops, and the two companies ran it on two linked Mac Studios, where it fixed bugs in a real app by itself. Meta's top model, Muse Wonder, stays closed. Engadget puts a working Mac setup at roughly $38,000, and Chinese free models still beat it at some coding.
- The claim
- Meta calls it the most capable American AI model anyone can download for free.
- Austin's take
- Austin is excited to see strong American open models, after a year when Chinese open models had no real US rivals. He notes that most people don't need the $38,000 hardware, because open weights can run on a rented private server or in the cloud for much less. He credits Alexandr Wang and Mark Zuckerberg's team.
- Why it matters
- A capable open model you can run yourself means your data never has to leave your building, which matters for privacy-sensitive work.
- What to do
- If data control matters to you, look at running it on a cloud server before buying hardware. Compare it against Chinese open models for your coding tasks.
Watch from 20:29
- What happened
- Accenture reportedly told staff it now covers only $29 a month of Claude after some employees burned through $1,000 of credits. Since October 1, anything over that must be billed to a project or team, and one worker said their access was cut off entirely. Workers told 404 Media they regularly spent $500 to $1,000 a month. Accenture sells AI advice to clients while limiting it for its own staff.
- The claim
- Accenture says it makes AI available to all its people with the right tool for each job.
- Austin's take
- Austin thinks $29 a month barely covers anything at API prices. He sees two readings: Accenture is cutting costs as consulting and implementation revenue face pressure, or its people aren't getting enough value, which would be bearish for the whole AI market. He personally gets far more than $29 of value, though his subscriptions are heavily subsidized. He expects major breakthroughs before the end of the year.
- Why it matters
- How heavy users like consulting firms budget for AI is an early signal of whether companies are really getting a return on these tools.
- What to do
- If your employer pays for AI, track what you actually get out of it so you can make the case for your budget. Watch what other big consulting firms do next.
- Not confirmed yet
- The policy is reported, based on what workers told 404 Media. Austin's reasons for the cap are speculation.
Watch from 26:30
- What happened
- OpenAI's update, GPT Live 1.1, adds voices in French, German, Hindi, Mexican Spanish, Korean, and Japanese. It started rolling out to every ChatGPT user on Thursday through the voice button, but the staged rollout means some people won't see it for a few days.
- The claim
- OpenAI says the update brings better accents and less 'checking' chatter while it searches.
- Austin's take
- Austin loves ChatGPT voice for brainstorming in the car but has to tell it every time to stop saying things like 'let me check on that,' so he is happiest about that fix. He thinks voice will become a main way people interact with their AI agents. He says he talks to AI less than half the time and uses Whisper Flow for dictation in a private office.
- Why it matters
- Non-English speakers get voices that sound local rather than translated, and conversations should feel less cluttered.
- What to do
- If you don't see the new voices yet, check again over the next few days. If filler talk bugs you, try the update before writing custom instructions to suppress it.
