>Jensen Huang gets ousted from Nvidia literally 12 hours ago
>OpenAI and Anthropic immediately form an alliance to back congress to ban AI chip exports to china
This was a targeted takedown.
Today we’re introducing Tofu-1, Bland’s speech-to-speech model.
The highest-stakes calls rarely happen in good conditions: from moving cars and hospital rooms, over weak signal, with several people talking at once.
Tofu-1 is built for those calls. It listens while it speaks, reasons while it talks, and can tell a caller apart from everything else on a line, like a TV or a crying baby.
Join the preview: bland.ai/speech-to-spee…
Safety evaluations are critical for understanding whether a model is ready to deploy, which safeguards to build for it, and where to focus training and research.
On our Engineering Blog, we offer 10 recommendations for how to design, run and report them.
Google recently changed its search results so they route through "GOTO" URLs to stop scrapers.
This created a bot problem for Brave Search and its users.
Let's talk about what happened, and how we solved it in a user-first way. 🧵
Personal Intelligence in Gemini securely connects information from apps you already use to make your experience more personalized and helpful. Today, we’re adding Google Wallet to the list of connected apps so you can get more tailored responses in the @GeminiApp around personal finance questions, based on your spending habits, including:
📈 Generate dynamic charts tracking cash flow, asset allocation, and budgets directly in your chat to help you visualize your spending.
📅 Find forgotten subscriptions and recurring charges across your linked accounts, and Gemini will also flag any price increases.
🎁 Set a budget for the holidays by asking “How much did I spend on gifts and travel last year? Use that to help me set a holiday budget” and Gemini will look at your transactions from the last year to suggest a budget tailored for you.
You’ll also have the option to securely link your financial accounts from your credit cards and banks to Wallet via @Plaid to unlock deeper insights.
Meet GR00T 2, our most advanced, open System One model for robotics research. It learns skills from human demonstrations and predicts how its actions will change the physical world.
Those skills can carry across different robot bodies, reducing the need to teach each robot every task from scratch.
Watch GR00T 2 put that research into motion. 🧵
This is a difficult moment. It’s hard not to look at the journey this company has taken and not be filled with respect for what Jensen built. The spirit of the company remains strong: "mission is the boss”.
That said, as NVIDIA finds its way forward, I want one thing to be clear: NVIDIA's contributions to open source are unwavering.
Jensen has been vocal about this, and that commitment runs deep in NVIDIA's people and culture. We all understand the importance of open source.
NVIDIA is a huge supporter of open source because we understand something.
If you want to diffuse technology, you do so through an ecosystem, not a single company or entity.
No one owns the internet, no one owns servers, and no one owns operating systems… but through open source, an ecosystem of closed and open innovation flourishes.
When we talk about delivering agents safely and securely, that too will take place because of open source. For example, ensuring people have the best models to secure themselves at machine speed unburdened by arbitrary guardrails decided behind closed doors.
Open source & AI are often discussed together, but it’s worth reminding, open source predates AI. It is a mechanism for sharing ideas inspired by the hippies at computer labs of California in the 1960s.
Before AI, the largest shift I lived through was the internet. It was remarkable seeing the fight for open source take place there. Between browsers, runtimes, net neutrality, etc. What came as a result was HTML5 and an extremely vibrant ecosystem that built the internet as we know it today.
NVIDIA and its partners hold an incredibly important role in delivering this technology. NVIDIA contributes to every layer of the stack. Runtimes, models, harnesses, programming languages, frameworks, etc. None of that changes.
Long live open source.
For my first post, I’m sharing a letter @NVIDIA signed on why open models matter.
AI will transform every industry, power every company, and be built by every country.
Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty.
The world needs both frontier closed models and frontier open models.
images.nvidia.com/pdf/Open-Weigh…
Starting with v1.98 on desktop, Brave removes hidden identifiers that Meta adds to your photos on Instagram and Facebook.
This makes it harder for Meta to link the same photo across accounts and sites.
Now your dot can securely complete purchases on your behalf—with your approval.
With ChatGPT Wallet, agents pay without access to your actual card details. Every purchase requires a security check.
ChatGPT Wallet is rolling out starting today to dots, for Pro subscribers in the US.
openai.com/index/introduc…
For the first time in a very long time, we have a non-OpenAI model as #1 on MathArena! Gemini 4 Carbon takes over the first place from GPT-6.1 Sol, with Fable 5.2 taking #2.
Omarchy really makes these RTX machines shine! Putting finishing touches on the out-of-the-box local AI experience now too. Crazy what's possible with the latest big jump in open-weight models.
For the past few weeks, I've been tuning the Omarchy experience for Nvidia RTX Spark.
Today is the first time just using it for gaming and I have to say; I'm impressed!
Performance in WoW is solid 60-ish FPS on High / Ultra settings, on battery, in the middle of Storm wind. 🤯
🔥 Hot metal. Blue fire. Stage 2 completed its mission duty cycle on the first try: a 400+ second burn and coast demo including an igniter hot fire, demonstrating everything it needs to do in flight.
What we shipped this week @google
Monday: Two new 8th-gen TPUs
Tuesday: Gemini 4 Carbon
- AlphaProof Code verified Signal protocol
- Gemini Robotics 2.5
Wednesday: AlphaGenome 2 for disease research
- Gemini Decisions API in AI Studio
Thursday: Gemini 4 Hydrogen
- intelligent eyewear pre-orders
Friday: U2 new album (ok not us but exciting!)
onwards!
For my first post, I’m sharing a letter @NVIDIA signed on why open models matter.
AI will transform every industry, power every company, and be built by every country.
Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty.
The world needs both frontier closed models and frontier open models.
images.nvidia.com/pdf/Open-Weigh…
In other bad news, it's been found that @bestbuy is selling uncensored laptops that could be used for cyber attacks. There are zero restrictions on these machines.
In other bad news for Nvidia: @stokel found that Hugging Face, which Nvidia recently agreed to acquire, appears to be hosting a load of models that could be used for cyber attacks.
transformernews.ai/p/hugging-face…
We killed $349/mo Semrush+Ahrefs💥
Run a SEO-optimized campaign in one prompt for just $0.61!
With aimlapi.com/mcp your agent can:
1. Analyze your website and competitors’ ads
2. Score 10k+ keywords with @TypesafeAI’s Jev
3. Create, set up and launch the Google Ads campaign
0% markup. No subscriptions. No extra signups.
First 100 commenters will get a $10 promo code to try it!
Give your agent this prompt: “Connect aimlapi.com/mcp and build a Google Ads campaign for my company: yourwebsite.com”
Developing a drug takes years, and most candidates never reach patients.
Amortizing the cost of failed drugs across stages, the present value cost of developing a drug is close to $2B. The process takes 13 years on average, with rising clinical trial costs putting further pressure on R&D productivity.
AI could improve those economics by increasing the probability of success and reducing development time, from identifying promising molecules to designing clinical trials and improving patient recruitment.
That may enable companies to pursue therapeutic targets and indications that might otherwise have been uneconomical.
Full piece on why AI will make pharma even bigger: a16z.news/p/big-pharma-c…
Register for Get Certified: Special Edition to earn a brand new badge!
Our hybrid Professional Agentic Architect (PAA) cert includes no-cost prep & hands-on Google Skills challenge labs. Google Cloud customer email required.
Apply by Dec 9 → goo.gle/4iMtM8U
Introducing Anywhere (YC F26): the missing map for your agent.
Let it see the world through your eyes.
Works with any AI, including ChatGPT and Instinct.
Try it here:
apps.apple.com/us/app/anywher…
I think @grx_xce is one of the most formidable founders of all time. The proof: she’s leading one of the fastest growing YC companies ever.
this video is worth a watch just to understand how her mind works.
Intelligence (S25) is building benchmarks for the parts of AI that are hardest to measure. Their first product, @DesignArena, lets people use frontier models to build websites, images, videos, and other creative work, then turns those real-world interactions into data that helps model labs understand where their models succeed and fail.
Design Arena has grown to more than 7 million users across 192 countries, and @Intelligence_ai reached a $60 million annualized run rate in about a year.
In this fireside, Intelligence founder and CEO @grx_xce joins YC’s @agupta to explain why measuring what people actually want could become increasingly important as AI models multiply and specialize. They also talk about what she’s learned building a new category at the frontier of AI, where moving quickly and learning as you go can matter more than already having the answers.
00:00 — How Design Arena Works
03:21 — From AI Games to Creative Benchmarks
06:43 — Measuring Human Taste at Scale
09:56 — A Marketplace for Specialized AI
13:52 — The Small Team Behind a $60M Run Rate
17:04 — Hiring Fast Learners and Taking Action
20:56 — Handling Rejection and Living in the Future
What if you could generate an entire product flow in just five seconds?
Early design is rapid and iterative. To amplify your creativity, you need a synchronous design partner that keeps up with your train of thought.
So we built a rapid design experiment in Stitch. It’s powered by the Gemini Decisions API to help explore what happens when early design iteration takes seconds instead of minutes ⚡️:
5️⃣ seconds from rough prompt to a multi-screen flow
2️⃣ seconds from tap to new variations
1️⃣ click to dive into Stitch for deeper UI polish
⚠️This is an early MVP to test how fast early ideation can feel, and we’ll be folding your favorite parts into the full Stitch experience soon.
Look for the banner in Stitch to try it yourself or see how it works below 🧵👇
Pharmaceutical companies are among the largest, most profitable, and most strategically important businesses in the world.
The top 30 pharma companies represent roughly $5T of market cap. Eli Lilly alone has a market capitalization of over $1T, putting it in a club historically dominated by tech darlings.
In Q3 2026 alone, Lilly's GLP-1 portfolio did close to $17B, nearly matching Anthropic's revenue in that quarter. Analysts expect Lilly's GLP-1 portfolio to grow to over $100B in annual sales by 2030.
Big Pharma Can Rival Big Software: a16z.news/p/big-pharma-c…
AsyncGRPOTrainer is now stable in TRL v1.18.0!
generation and training run at the same time on separate GPUs, instead of taking turns
we wrote up how to run it, what you can train with it and how it compares with GRPOTrainer: huggingface.co/spaces/trl-lib…
Today, @bayesiankalki and I are launching Mnemonic Labs (YC F26). We're a neolab training a foundation model for quant trading that learns directly from market activity. In <6 weeks we’ve built our training pipeline and model architecture. Our current training run uses over 25TB of tokens.
We met 28 years ago shortly after I was born (we’re brothers!). We’re Olympiad finalists (5x USACO combined) and MIT graduates. Kalki made $20M in annual PnL at 22 at Tower Research and I was the youngest quant at Citadel at 19. Over the past ten years, we’ve been part of quant trading’s evolution from linear models to trees and now neural networks.
Learn more about us at ycombinator.com/launches/XRK-m… or reach out at founders@mnemoniclabs.ai!
Construction still runs on manual work and tools that weren't built for AI.
Today we're launching Constor (YC F26), the AI-native OS for construction. Starting with takeoffs: days of work, done in minutes.
s/o @elioisbuilding & Louis Pires
Wiz co-founder and CTO Ami Luttwak says cyberattacks on startups can happen faster than ever:
"I work with the CEOs of startups that are getting attacked. Let's just be in their shoes. You wake up, and your security team says, 'Something bad happened overnight. We tried to stop it, but they got in.'"
"The entire attack, again, takes 20 minutes... from the time they get in until they take all of the secrets, all of the AI keys of all of the customers. And now you have to figure out what are you going to say to all of the customers, and what you're going to do."
"But if I'm a startup, especially if I have anything attackers want... I am on their target. It doesn't have to be a big company."
"If you have AI keys, probably better to put a limit on them, because if they steal them, you don't wanna get a lot of charges, which we see happen a lot, and it's really, really sad. But also make sure that if you have a target on your back, you do need to put a lot of focus on security."
@amiluttwak
Wiz co-founder and CTO Ami Luttwak says AI can now carry out cyberattacks in 20 minutes that once took skilled attackers weeks or months.
In his view, companies need to continuously attack their own systems with the latest AI models to stay secure. For startups, that means building automated testing and patching into the development process from the beginning.
Ami joins a16z's Sarah Wang and Zane Lackey to discuss what changed during Black October, why patching is falling behind, the cost of continuous security testing, and why better security architecture matters more than simply buying more tools.
0:00 Intro
1:04 Welcome to Black October
3:19 How AI made every day a zero day
6:51 The PHP vulnerability: 70% of the internet
9:19 Why patching can't keep up
11:32 How Wiz uses AI to find vulnerabilities
16:59 The cost of continuous testing
19:23 Cloud security in the AI era
22:18 The red agent: attack yourself first
26:41 Agent swarms attacking your own systems
28:43 What founders must do before they ship
32:15 Where security budgets should go
36:10 Most of what security teams do isn't security
43:34 Scan for Good and public infrastructure
YouTube: youtu.be/PbD9pVXxsYM@amiluttwak@sarahdingwang@zanelackey
A massive week for AI infrastructure!
We just released our newest TPU systems, open software contributions, and orchestration tools on Google Cloud.
Here is a breakdown of what we shipped ↓
In October, we launched new ways to work with Copilot:
• Copilot automations for issue triage
• Projects in the GitHub Copilot app for work across multiple repos (or without a repo!)
• Copilot Fusion for routing to frontier models
Find out how our dev team is using them ▶️
Exa now runs in Europe.
Our index and retrieval, rebuilt in Stockholm. Requests are stored and processed in the EU.
Available today for enterprise customers.
We rolled out Gemini 4 Hydrogen inside Google @Antigravity alongside a new Orchestrator mode in the model dropdown: Orchestrator (Carbon) and Orchestrator (Argon). 🪄
Running a heavy frontier model on every small code edit slows you down and burns quota. With Orchestrator, @Antigravity automatically pairs Gemini 4 Carbon or Argon with Gemini 4 Hydrogen in the same session for an optimized balance of quality, efficiency, and latency:
— Carbon or Argon explores dependencies and builds the plan
— Hydrogen quickly writes the code, updates packages, and runs tests
— Carbon or Argon steps back in to adjust the plan if a test fails
Learn more: goo.gle/3UxMLMx
Yesterday we announced the establishment of the new Center for Mathematics and AI at Courant (MAIC). Its mission is to make frontier AI systems an object of scientific inquiry, one that the broad mathematical community can dissect, improve, reproduce—and ultimately—understand.
The story as I understand it seems to be that Nvidia legally exported chips to Asian data centers legally allowed to buy them. Chinese nationals then legally leased time on those chips over the internet.
This embarrassed a bunch of bureaucrats who thought their abuse of laws designed to prevent unauthorized export of weapons to control microchips was getting something done, and they tried to cover up their foolishness with an ex post facto witch hunt.
The hacker claims to have obtained 6.9GB of data containing 46,000 employee records, and is asking for 3 BTC ransom, or the information will be released.
Today, I am launching @ninesrobotics with @tianx2li.
We build autonomous robots for AI data centers.
Currently, AI only optimizes itself in software. By building the automation layer for server rooms, we are creating the world’s first fully software-defined, physically adaptive AI factory - the essential bridge that allows agentic AI to autonomously manage, manipulate, and optimize its own physical hardware stack.
🛡️ Brave strips hidden Meta identifiers from Facebook and Instagram photos
Brave added a desktop privacy feature that automatically strips hidden identifiers Meta embeds in Facebook and Instagram photos.
It removes the marker when you save a tagged image or upload one to a website, reducing the risk the file can be used to link activity across accounts.
Details ↓
🚀 @GoogleAI's Gemini 4 Hydrogen is now generally available in GitHub Copilot.
Designed for fast, efficient agentic coding, tool use, and multistep workflows, it has improved instruction compared to Gemini 3.8 Flash.
Find it in the GitHub Copilot app, CLI, and @code.
"Every org, every individual is going to have a security incident in the next year."
Pablo Sabbatella, Opsek founder and Security Alliance member, explains on a new #TheBraveTechnologist episode why AI is making security worse before it gets better: brave.com/podcast/e146/
The Derp State is out to ban open weight models, cartelize labs with frontier models, make sure no one can go off the grid. This is as dumb as it sounds, it can't work. But they will try, because they're stupid and evil.
You sort of set him up, didn’t you, by creating the impression that senior government officials anonymously provided information about spying, when in fact the info came from a lobbyist at a competing AI company?
What we shipped this week @google
Monday: Two new 8th-gen TPUs
Tuesday: Gemini 4 Carbon
- AlphaProof Code verified Signal protocol
- Gemini Robotics 2.5
Wednesday: AlphaGenome 2 for disease research
- Gemini Decisions API in AI Studio
Thursday: Gemini 4 Hydrogen
- intelligent eyewear pre-orders
Friday: U2 new album (ok not us but exciting!)
onwards!
The @FAAnews issued a nationwide ground stop for all U.S. airports following a communication issue impacting all Air Route Traffic Control Centers. FAA technicians are investigating the issue. We will provide immediate updates as we have them.
From MAX POWER to Mars, it was a week to remember at @NASA!
🇺🇸 MAX POWER takes over @NASAKennedy
🌕 New Moon Base awards
✈️ X-70 Fractal unveiled
🧑🚀 Astronaut recruitment opened
🔴CHAPEA Mars simulation concludes
🛰️ ESCAPADE sets off to Mars
Here’s your NASA Minute!
blurry photo in the field? Jev makes a snap judgment to find out.
@elise_ai clocked Jev at 94ms a photo: ~30x faster than Sol, with ~50x as many judgments for the same cost
One of the most important Jev patterns to learn is what we sometimes call "speculative questions".
In the majority of cases, it's faster and even often cheaper to batch questions together in to a single call, *even when you don't yet know if you'll need the answers!*
Why? Many reasons:
Speed! Oftentimes you can accomplish the entire task in a single call, instead of multiple sequential calls in different logic paths.
Cost! You only pay for tokens in state once per call, so batching questions also avoids paying for the state multiple times. When your state is large, batching questions often saves cost compared to sequential calls that duplicate state.
Questions never affect the output of other questions in the same call, so accuracy will not suffer.
If you listen to one podcast this week, listen to this one.
As the CTO/Co-founder of Wiz, CTO of MSFT Cloud Security, CTO/Co-founder of Adallom, @amiluttwak has spent his entire career finding what others miss...and building what comes next.
Ami unpacks Black October, the explosion of vulnerabilities, AI attacks on AI infra, agent swarms, and why + how the entire security industry needs to change.
@zanelackey
Wiz co-founder and CTO Ami Luttwak says AI can now carry out cyberattacks in 20 minutes that once took skilled attackers weeks or months.
In his view, companies need to continuously attack their own systems with the latest AI models to stay secure. For startups, that means building automated testing and patching into the development process from the beginning.
Ami joins a16z's Sarah Wang and Zane Lackey to discuss what changed during Black October, why patching is falling behind, the cost of continuous security testing, and why better security architecture matters more than simply buying more tools.
0:00 Intro
1:04 Welcome to Black October
3:19 How AI made every day a zero day
6:51 The PHP vulnerability: 70% of the internet
9:19 Why patching can't keep up
11:32 How Wiz uses AI to find vulnerabilities
16:59 The cost of continuous testing
19:23 Cloud security in the AI era
22:18 The red agent: attack yourself first
26:41 Agent swarms attacking your own systems
28:43 What founders must do before they ship
32:15 Where security budgets should go
36:10 Most of what security teams do isn't security
43:34 Scan for Good and public infrastructure
YouTube: youtu.be/PbD9pVXxsYM@amiluttwak@sarahdingwang@zanelackey
This is a fascinating story. David started building a platform that could decompile games and make mashups and mods, which went pretty viral.
But then he realized that same platform could make ~infinite customized worlds for agents and robot sims to do RL. Excited to see this launch.
introducing the mireye agent!
real-world research means digging through property records, government databases, maps, and spreadsheets.
now mireye does the research for you.
the agent has access to 160m+ u.s. parcels, 600+ data fields, and proprietary databases.
in one conversation:
- research ownership, zoning, property values, and demographics.
- check infrastructure, nearby businesses, and environmental risks.
- enrich properties, calculate distances, and compare locations.
available through our chat interface, api, and mcp.
the best part? no extra charge for the agent.
give it a spin → mireye.com
We built these tools with security and privacy at the center. If you choose to connect your Wallet, Gemini can only read your accounts; it can’t move, manage, or invest your money. Learn more.
A working exploit once took a professional a median of 22 days to develop. Recent AI-assisted examples have taken hours.
@amiluttwak: " If you have AI keys, it's probably better to put a limit on them... If you have a target on your back, you do need to put a lot of focus on security."
Wiz co-founder and CTO Ami Luttwak says cyberattacks on startups can happen faster than ever:
"I work with the CEOs of startups that are getting attacked. Let's just be in their shoes. You wake up, and your security team says, 'Something bad happened overnight. We tried to stop it, but they got in.'"
"The entire attack, again, takes 20 minutes... from the time they get in until they take all of the secrets, all of the AI keys of all of the customers. And now you have to figure out what are you going to say to all of the customers, and what you're going to do."
"But if I'm a startup, especially if I have anything attackers want... I am on their target. It doesn't have to be a big company."
"If you have AI keys, probably better to put a limit on them, because if they steal them, you don't wanna get a lot of charges, which we see happen a lot, and it's really, really sad. But also make sure that if you have a target on your back, you do need to put a lot of focus on security."
@amiluttwak
"Notion says Fireworks retained data despite the zero-data-retention clause!"
This goes for any market but I think this is especially true for AI and VPNs, ZDR is a pinky promise you shouldn't trust at all.
You can now use Google's Gemini Embedding 2 with Pinecone integrated embedding.
Set gemini-embedding-2 on a text field and send plain text. That field can serve full-text search too, so hybrid search doesn't need a second index or embedding pipeline.
🔗 pinecone.io/blog/integrate…
I've been wearing my v0 Vexo prototype every day for the past month or so.
The coolest use case for me - which they didn't really mention in the video - is when it passively picks up things I need to do, buy, or note without me doing anything. Those experiences feel magical.
E.g. I met 20 people at an event last week - and later that night my agent gave me a rundown of everyone I talked to, mapped their name to what we talked about, and made it easy for me to stay connected.
Congrats on the launch guys!
Backed by YC and Paul Graham, we're building Vexo, an always-listening AI bracelet that turns your conversations into action.
Built by engineers from Tesla, AMD, and Amazon.
One sentence to Vexo can:
•Get Claude building the feature you wanted
•Text the family
•Block your calendar for dinner
•Order the groceries
•Find the founder you met at last week's event
Every conversation adds context, so you explain less and get more done.
Get yours now, shipping start January 30th 2027
vexoai.com/checkout
Swing by Northwestern University, Ruan Conference room (lower level), at 11am Central! Will be speaking about end-to-end learning of predictive abstractions on any data! We will see how and why humans got so good at this, and why JEPAs are the next step
statistics.northwestern.edu/events/seminar…
The investigation into FLI's crypto donation is now updated with new information from the Bitcoin Policy Institute's report:
How FTX and Alameda Research may have helped turn the enormous but largely illiquid SHIB donation into hundreds of millions of dollars for FLI.
Backed by YC and Paul Graham, we're building Vexo, an always-listening AI bracelet that turns your conversations into action.
Built by engineers from Tesla, AMD, and Amazon.
One sentence to Vexo can:
•Get Claude building the feature you wanted
•Text the family
•Block your calendar for dinner
•Order the groceries
•Find the founder you met at last week's event
Every conversation adds context, so you explain less and get more done.
Get yours now, shipping start January 30th 2027
vexoai.com/checkout
One of the more ambitious teams I'm working with this fall.
It turns out batteries are perfect for rail: they're easy to add to existing locomotives, they're much cheaper than diesel, and it takes less time to swap them than it does to change a crew.
Big day for me: @TagessonEmil and I are launching Ailit, swappable batteries for freight trains 🚂🔋
We met at KTH in Stockholm with the same conviction: the future is electric.
North American railroads burn ~$15B of diesel a year. But their locomotives already run on electric motors.
So we built the world's largest swappable battery packs to power them directly.
→ Packs ride behind the locomotives railroads already own
→ Swap in minutes at stops trains already make
→ Railroads pay per kWh, priced below diesel, nothing upfront
We just finished Superwagon, our first full-scale prototype. Next, we roll it out across our energy network.
Rail ran on steam. Then on diesel. We're building what comes next.
Backed by @ycombinator.
🚀 Microsoft Ignite is almost here, and we have a challenge for you!
Can you get Basket IQ, an AI shopping agent, launch-ready on Azure HorizonDB? Explore the database with the PostgreSQL extension for @code and put your engineering skills to the test!
🎮 Play online or at Microsoft Ignite in the Hub starting Nov 17:
🔗 aka.ms/decision-point…
Excited to announce our research partnership with the UK @metoffice. Their climate science expertise has been critical to evaluating Google’s WeatherNext models.
Looking forward to continuing to work together with their scientists and our teams at @GoogleDeepMind and @GoogleResearch to advance AI for weather forecasting.
Working with @GoogleDeepMind, Met Office scientists will be reviewing AI model WeatherNext-3. The model analyses existing data and weather patterns for insight that complements our physics-based approach.
Read more here: metoffice.gov.uk/blog/2026/expl…
Big day for me: @TagessonEmil and I are launching Ailit, swappable batteries for freight trains 🚂🔋
We met at KTH in Stockholm with the same conviction: the future is electric.
North American railroads burn ~$15B of diesel a year. But their locomotives already run on electric motors.
So we built the world's largest swappable battery packs to power them directly.
→ Packs ride behind the locomotives railroads already own
→ Swap in minutes at stops trains already make
→ Railroads pay per kWh, priced below diesel, nothing upfront
We just finished Superwagon, our first full-scale prototype. Next, we roll it out across our energy network.
Rail ran on steam. Then on diesel. We're building what comes next.
Backed by @ycombinator.
I had a great conversation with @tdechant at the @SOSV Deep Tech Summit about a lesson from @ycombinator that still shapes how we work at @Helion_Energy : talk to your customers and listen closely to what they need.
In fusion, it’s easy to spend all our time talking about the physics behind our work, but it’s conversations with customers that help ensure we are building fusion generators that truly address their real-world energy needs.
Thanks for the conversation, Tim!
We've published a guide to building an open AI-native stack on Kubernetes for #KubeCon#CloudNativeCon.
Scale from prototype wrappers to multi-host model serving, dynamic GPU/TPU allocation, and production agent orchestration ↓
Pipecat 1.16 adds PerceptionReader, something I'm really excited about. How someone speaks can matter just as much as what they say, and voice agents have mostly been missing that.
We’ve been testing it with Google Decision 1, released yesterday. Stream it audio and you get a pretty accurate read on the prosody, tone and sarcasm of what the user said.
Try talking to it at this demo here: …pecat-google-decision-demo.vercel.app
We've exploring many different ways of measuring expressivity... more to share on that soon!
➡️ Pipecat Docs: docs.pipecat.ai/api-reference/…
➡️ Google Decision 1: blog.google/innovation-and…
So instead of a query and then scrape the SERP for links, bot clicks on links, redirects to "real" page.
Could be stale results! Google answered "midterms" with 2022 not 2026 results the other week, to suspected scraper queries.
Game's afoot, but G did raise costs on scrapers.
Cloud Pathways is now generally available.
This is the distributed ML runtime originally built by @GoogleDeepMind to train Gemini, now available to cloud customers.
Using a single-controller architecture, it orchestrates large-scale TPU systems across multiple slices as if they were one machine ↓
Google did raise costs recently, substantially. First, they started giving suspected scrapers stale results (see ). Then they did theyr ?goto= thing, which requires scrapers to simulate a click and load the redirected-to page. That lets Google analyze them.
Ahrefs, SEMRush and ALL the rank trackers seem to be bugging out BIG TIME right now vs. actual SERPs 🚨
Looks like Google could be implementing ANOTHER anti-scraping measure after they lost the SERPAPI court case.
It seems to just fill the entire first page with UGC content or YouTube videos instead of actually surfacing the real ranked pages.
I don't personally believe it's anything to do with spammers or hackers, it's more to do with them trying to protect their own scraped data (I know, ironic, isn't it), AI Overviews/AI Mode etc..
And I know it's been a pain fixing all of this for the SEO and rank tracking companies anyway, but looks like there's a lot more hurdles in front of them to come.
