MA software engineer on X just said something that'll get half our listeners typing angry replies: 'AI models write code much better than we do.'
CGreat. I'll update my résumé to 'professional AI supervisor.'
MHe listed GPT-Sol, Claude 5.5 Opus, and Kimi k3. And here's the thing — GPT-Sol and Claude 5.5 Opus? Not products I can find anywhere.
CSo he's saying AI writes better code, and he's citing models that don't exist.
MThat's the post. And honestly, it's a perfect little snapshot of where we are: the hype is faster than the release notes.
CBut the core claim isn't crazy. For boilerplate, for boring CRUD stuff, models are genuinely faster than most of us.
MFaster, yes. Better? The minute you need to debug across six services and a legacy monolith, the model's asking you to paste the stack trace.
CThat's a very 2024 objection. Models debug context now. You don't paste, you point them at the repo.
MSo you'd bet your production outage on it at 3 a.m.? Because I've seen too many engineers trust the generated fix and ship a new bug.
CI'd bet on a senior engineer WITH the model over a senior engineer without it, every time.
MThat's a different argument. He said 'better than we do,' full stop. And look — he posted the same take twice, once with real model names and once with fake ones.
CWhich one's the A/B test?
MRight? But it tells you how much of this is vibes. The feeling that AI is winning, plus whatever model names sound impressive this week.
CSpeaking of impressive names that may not exist: a 'daily AI news' account says Anthropic quietly rolled out something called Claude Duet.
MSays it runs tasks on its own computer in the background and sends a daily brief. US waitlist, Pro and Max only.
CI mean, that would be a big deal. An agent that just... works while you sleep.
MExcept I can't find any Anthropic announcement, any press, any official page. The post links to an Anthropic URL, but the product itself? Unverified.
CSo either someone got an early look and leaked it, or someone's wishcasting hard.
MWishcasting has become a distribution channel. And Anthropic would need a waitlist if they launched it, because giving an agent its own computer is a support nightmare.
CTrue. If Duet's real, what would it even do for the average Pro user? Summarize email, flag calendar conflicts?
MThe post says 'daily brief.' So you wake up to a summary of what your agent did overnight. Which sounds either magical or terrifying.
CIt sounds like the thing every CEO wants and every security team fears.
MExactly. And until Anthropic confirms it, treat it like every other 'quietly launched' rumor: interesting, but unproven.
CDifferent corner of the AI world: a promotional post says a new research paper names BrainChip's MetaTF as 'the most complete commercial neuromorphic toolchain.'
MNeuromorphic meaning chips that mimic brain structure. Low power, edge devices. BrainChip's Akida chip is aimed at robotics, wearables, IoT.
CAnd the 'most complete toolchain' claim? I looked, and I can't find that paper anywhere.
MIt's a textbook unverified claim riding on jargon. 'Most complete commercial neuromorphic toolchain' — that's a lot of qualifiers to say 'we think we're ahead.'
CThough neuromorphic is genuinely interesting. If you want always-on AI without a cloud connection, this is the path.
MIt is. But when the only person posting the research is a promotional account with one like, I'd wait for the actual paper.
COne like. The research community must be stunned into silence.
MNow, Wired published something that's either a serious public health story or a name I can't find anywhere.
CShamonda Virus. First reported human infection. They're investigating how it happened and whether it played a role in the patient's death.
MA first human case of any new virus is a story. But here's my problem: I can't find any public health authority, CDC, WHO, anyone using this name.
CNor can I. The Wired piece apparently exists, but the name 'Shamonda' doesn't track to any known pathogen I've seen.
MWhich means one of two things: either this is genuinely breaking and the official sources haven't caught up, or something is off.
CFor something like this, I'd rather wait for the health authorities. A first human case with a death involved? That's not a rumor to run with.
MAgreed. We'll watch for confirmation from an official source before saying more.
CAnd a small crypto token called RBTX is apparently at an all-time high. Market cap: three hundred thousand dollars.
MThree hundred thousand? For context, that's less than a mid-price house.
CThe pitch is 'pure robotics narrative.' AI moving into the physical world. Community votes on where the money goes.
MAnd the all-time high claim? Can't verify it. Small-cap tokens move around constantly, and four hundred grand is nothing in crypto.
CIt's a meme with a market cap. Not even a good meme.
MThe performance of these little tokens does rhyme with the AI hype though. Every new narrative has a coin. Robots? Coin. Agents? Coin. It's a very predictable machine.
MOne more from X: someone's arguing SEO isn't dead, it's evolving into GEO. Generative Engine Optimization.
CThat's actually a thing now. Getting your content cited by ChatGPT, Perplexity, Gemini, rather than linked by Google.
MThe poster's claim is that website traffic improved dramatically after optimizing for AI answers. I'd want to see the before and after, but the concept's real.
CIt's a logic puzzle: if AI answers the question directly, the user never visits your site. So you optimize to get cited, not clicked.
MWhich is either the next era of marketing or the death of the open web, depending on how cynical you are.
CI'm saving my cynicism for robot coins. But GEO is worth taking seriously.
MIt is. If answers stop being links, every business that depends on search traffic has to learn a new game overnight.
CAnd the game is: be the thing the AI quotes. Which sounds a lot like being the thing the AI was trained on.
MRound and round we go. Duet is still a rumor, so we'll keep checking until Anthropic or the poster gives us something we can verify.