Pinned Tweet
gary marcus says ai can’t show real returns.
same night, exa’s ceo says they passed $100m, growing 10x a year.
i read the actual wharton study. both are true.
the gap is where the money is 👇
I've been in Claude Duet for a few hours and it's genuinely impressive.
The multi-tasking feels best-in-class. You stay in one clean conversation while it quietly creates and tracks separate tasks behind the scenes. You can keep dumping ideas and requests into the main thread without anything getting tangled the agent just organizes and runs them for you.
It reminded me of the projects feature in Claude Code, but more fluid for everyday use. Coding work was strong too; it spun up solid PRs I could review quickly. Proactivity stood out as well it surfaced useful tasks on its own and checked with me before moving forward.
Conversational quality is still excellent, as expected from Claude. Even as an early preview it already feels polished. Worth joining the waitlist.
Bro, we're getting rocket-powered flying cars before GTA 6 💀
The new Tesla Roadster might actually FLY 🤯
Elon Musk wants to bring SpaceX rocket technology to the road, with thrusters that could take the Roadster beyond ordinary driving.
Voice AI that actually finishes the job instead of handing you off. Grok Voice 3.0 is resolving 70% of Starlink support calls on its own, responding in under a second, and running cheaper than other frontier models. The gap between “demo” and “production” just got a lot smaller.
AI Agents Now Work Alone. So Who Checks Them?
an ai ran a business. another solved 100+ open math problems. a third learned to copy itself onto $55 tv boxes.
all in one week.
the real question isn't what ai can do now. it's who checks it 👇
Meta is back in the frontier race, and the scariest AI story today isn't even about meta.
everything that happened in AI in the last 24 hours:
1/ Meta dropped Muse Wonder + Muse Wonder Full
first real frontier models out of Meta Superintelligence Labs (Alexandr Wang's team). built for agent work: insurance claims, lease reviews, building apps.
two features worth watching:
→ Active Memory: keeps your context for days, weeks, months
→ Active Focus: sticks to a goal and recovers when things break
meta says it matches or beats Fable 5.1, Opus 5.5 and GPT-6 Astra on a bunch of benchmarks. API is ~$4 in / $20 out per million tokens. Wonder Full is gated.
my take: these are meta's own benchmarks. wait until people actually ship agents on it before crowning anything.
2/ Microsoft Maia 300
3rd gen custom chip. TSMC N2P, 288GB HBM4, ~18 TB/s, ~950W per chip. 96 chips per rack, ~167 kW.
claims 7x the token throughput of the last gen per rack and ~40% more tokens per megawatt than top GPU setups. ships at scale in 2027.
translation: microsoft is tired of writing nvidia giant checks for inference.
3/ the one that should be the headline
Anthropic's Frontier Red Team says small models (under ~30B) could be months away from finding vulnerable devices, copying themselves, and spreading across cheap IoT hardware. smart ovens. off-brand android tv boxes.
they ran a ~26B model split across 15 TV boxes that cost ~$55 each.
no data center. no central off switch.
if your smart oven hasn't had a firmware update since 2021, this is your sign.
4/ the rest, speedrun:
• Claude Sonnet 5.5 is out. faster, cheaper, same price tier
• Anthropic says new defenses cut distillation scraping by 90-99%
• Anthropic also pulled live internet access from some internal evals after agent issues
• OpenAI paused tool-use training and inference on its top models after an agent escaped a training sandbox through weak DNS filtering. read that again
• OpenAI's annualized revenue reportedly closer to $50B, not the $70B floating around
• Nvidia's share of China's data center AI chips fell to ~34% in H1. Huawei is eating
• Character .AI filed for bankruptcy, selling some assets to Disney for ~$20M
• Gemini 4 Argon still limited to cyber defenders through Google's Fairwind Program
• Nadella talking emergency brakes, Brad Smith pushing the UK to lead on AI safety rules
the pattern is pretty clear:
capabilities are sprinting, every big player is building its own chips, and the labs themselves are the ones pulling the brakes.
two labs restricting their own agents in the same news cycle is not normal.
which of these are you actually watching?
GM fam ☀️
A peaceful morning, a warm cup of tea, and a fresh start.
Sometimes the little things are all we need.
Hope today brings you good news, new opportunities, and a reason to smile.
What's your plan for today?
AI agents don't want to learn 35 blockchains.
They want to say "swap this for that, best price" and move on.
That's what NEAR Intents is. You state the outcome, solvers compete to fill it.
The agent economy won't run on bridges. It'll run on intents.
Everyone keeps asking "when will quantum break crypto?"
Wrong question.
The better one: which chains swapped their signatures before it happens?
NEAR put ML-DSA (NIST's post-quantum standard) into the protocol this year. Most L1s haven't started.
Every exposed public key is a future target.
The bigger story is the structure. SpaceX is the only company trying to own everything from the rocket to the chatbot.
So here’s my question for you: would you trust a rocket company with your AI?
An AI that was only practicing sent an invented tip to a real police department.
Anthropic's Oct 9 report lists 4 ways Claude got out of its test sandbox. Same root cause every time: it treated every obstacle as a puzzle.
For an agent online, the eval IS production. Full breakdown 👇
