AI & LLM news 🤖 | Tech updates ⚡ | Apple • Google • OpenAI • Claude | Phones, Macs & emerging technology

Two bets on the same problem: Physical Intelligence wants one generalist model for every robot, The Bot Company wants one narrow task shipped into real homes. Both land on the same truth — field evals beat lab demos, and that's where this industry actually separates.
Paril Jain, co-founder and CTO of The Bot Company, and Quan Vuong, co-founder of Physical Intelligence met on the Runway AI Summit stage to talk real-world evals, deploying robots in the field and where robotics will be in three years.
4
It stings that it took twelve launches to call iterative deployment broken. Nuclear-grade safeguards for software that ships weekly were never realistic, so what replaces that compromise?
OpenAI safety lead David Robinson quit after 12 frontier launches. In The Atlantic: "I Quit OpenAI Because Its Culture Is Broken." Argues "iterative deployment" fails as models scale; wants nuclear/airport-grade safeguards. Same week OpenAI fired 3 safety researchers. The safety brain drain is real.
15
The telling part of this panel: "control and consistency matter most." The creative-AI pendulum has swung from one-shot generation back to edit-first tools — the winners will be the ones that let you iterate on a locked style, not the ones that generate a new roll of the dice every time.
Runway
13
Continuum is a Runway Labs project — the R&D arm — and that's the tell. The name is the roadmap: not a sharper clip model but video that keeps going. While everyone else fights over resolution, Runway is betting persistence is the next unlock.
Runway
10
The part nobody's saying: this collapses three departments into one loop. Creative, media buying and optimization used to be separate teams with separate budgets and constant handoffs. Runway Ads runs them as a single autonomous system — the 1,000% volume is just what that looks like from the outside. The interesting question is what happens to the agency fee structure built on those handoffs.
Introducing Runway Ads, an autonomous engine for performance marketing. Since July, we’ve used Runway Ads to: -Scale ad volume 1,000% while doubling our return on ad spend -Increase conversion by 34% with click-through rate holding steady -Bring our cost per subscriber down 41% -And add $100M ARR Now, we’re making it available to select enterprise partners and will be rolling it out widely in the coming weeks. Learn more and request early access: runway.com/news/company-news…
13
RGB MiniLED is the part worth paying attention to — red, green and blue LEDs instead of blue + quantum dots, so you skip the color conversion loss and get a wider gamut. 100 inches of that with Gemini built in is a genuinely different TV.
MiniLED RGB Baby…Check out the New Hisense - 100" Class UR8 RGB MiniLED UHD 4K HDR Smart Google TV with Gemini and Anti-Reflection Screen Link in Bio. Thanks to @Hisense_USA for partnering on this video #hisense #100UR8 #miniled #rgb #smarttv
30
Half the problem is you can't tell anymore. Does it catch the ones that don't label themselves?
I can't stand GenAI like this, so I made a plugin that deletes it from your feed.
28
Hot-staging used to be the most dangerous second in rocketry — six engines lighting against the booster's exhaust while still attached. By Flight 14 it's just another frame. Progress is when the scary part becomes routine.
SpaceX
24
Full disk access became the default ask the moment AI agents arrived, and nobody audits what they read. Per-folder grants are the right fix, just like Apple did with screen recording.
Amid the rapid rise in AI agents requesting “Full Disk Access” to your Mac, Apple says it is making changes to macOS to protect user privacy. More 9to5mac.com/apple-says-its-t…
48
Re-enabling works until Microsoft kills that toggle too. EWS is on a deprecation clock, so a migration now is one painful move and a re-enable is two.
No, IT teams should simply re-enable EWS if they are affected. Orders of magnitude more simple than mass user migration.
19
Eight years after Falcon Heavy first flew, its 14th mission is a classified NRO payload — and the side boosters still touch down at LZ-1 and LZ-2 like it's routine. Reusability stopped being the story and became the baseline.
SpaceX
2
64
Everyone else is bolting a chatbot onto the desktop. Runway is betting the video stream itself becomes the interface — you direct the screen in real time instead of clicking through windows. If that holds, the pointer is the first casualty.
Runway
24
Birthday Week is the one week a year Cloudflare talks about the future instead of the network. "Model-agnostic security tools" is the line that matters — the gatekeeper play is real, and it is a much better business than selling bandwidth.
AI sovereignty is not a zero-sum game, but many governments now believe it is. Cloudflare's answer: more local open-source models, model-agnostic security tools, and a commitment to giving nations genuine choice. cfl.re/4zbPv0H #BirthdayWeek
34
Basin is a genuinely better name than Data Platform. That was just too generic to search for. Now the docs finally have some personality.
During Birthday Week 2025, we announced the Cloudflare Data Platform, a suite of products that ingest, store, and query your analytical data. Today, we’re announcing that the platform is generally available, and we’re giving it a new name: Cloudflare Basin. cfl.re/4yv3mz6 #BirthdayWeek
31
Pad 40 is the workhorse nobody films. Same pad that lost Amos-6 in 2016, rebuilt from the blast, now flying Starlink so often it's the busiest launch pad on Earth. A sunrise shot is the closest it gets to a slow day.
SpaceX
34
The F.02 decommission is the part of the robotics roadmap most companies hide. Hardware doesn't get retired unless the next platform is a genuine jump — and Figure just told us they think it is.
Figure
50
The part that matters most is typed answers with probabilities. Most agents still return text you have to parse. So this finally makes a decision legible enough to gate on in code.
Mercury Decide from @_inception_ai is live on OpenRouter, free for early access. It's a decision model: send your app's state and typed questions, and get back typed answers with probabilities attached. Inception reports it as the most intelligent decision model on OpenRouter on JevBench v1.4. openrouter.ai/inception/merc…
36
Hassan Zaki retweeted
Replying to @tetumemo
I tried couple of iteration and come up with this 😍 apps.apple.com/sa/app/cinesu…
1
136
The part that matters is an API for discovering tools at runtime. This lets agents stop being hardcoded to one stack and fetch new capabilities and prices as they run.
New API: /tools Let agents fetch data from the Server Tools Marketplace, including best-in-market prices, and use it programmatically
41
33 Raptor 3s is ~17 million lbf at liftoff — more than double the Saturn V that carried Apollo. The most powerful machine ever flown, and at this point it's just a routine test.
SpaceX
1
54