Jerene Yang

Martin Eberhard played a key role early at Tesla and even brought in Elon Musk. Then he was pushed out of the company for pennies. Founders get most of the attention in Silicon Valley, but the ecosystem is built on the people who take a chance and join early stage ventures. Unfortunately, those same people are often the easiest to cut out. Founders can dilute early employees, change vesting terms, and restructure equity with little effort and zero consequences (you can literally do it in Carta in 5 minutes). Daniel Kottke was a close friend of Steve Jobs and helped build the Apple I and II computers. He got no equity (Steve Wozniak eventually gave him a few of his own shares). Brian Acton co-founded WhatsApp and stayed through the Facebook acquisition. When he left Facebook over privacy concerns, he had to walk away from $850 million in unvested stock. There’s more awareness now around how vesting works, what happens during a liquidity event, and how to think about re-ups and dilution. That’s good. But the truth is, if someone really wants to screw you, they can. It still comes down to one thing: trust the people you work with. Or don’t work with them at all.

84

8 Comments

AMI Labs, co-founded by Yann LeCun, just raised $1.03B at a $3.5B valuation. I'm personally glad to have a small contribution to this venture. The thing I love most is the narrative the team isn't afraid to say out loud: "It's not your typical applied AI startup that can release a product in three months, have revenue in six months and make $10M ARR in 12 months." The most interesting things happening in AI right now don't fit the familiar venture path. They're not announced in press releases. They're built over years, behind the scenes. This team is going after something structural: a different architecture, a different premise, betting on what LLMs simply aren't fit for. The first wave of frontier AI was scale: bigger models, more data, more compute. The next wave will be about architecture, infrastructure, and the harder question underneath it all: are we even going in the right direction? More labs and researchers are now betting we didn't build this market in the most optimal way. That's a meaningful signal. That actually intersects directly with where we're going at Mirai Tech Inc. Different bets, same direction. Congrats Yann LeCun, Alex LeBrun, and the whole AMI - Advanced Machine Intelligence Labs team. This one's huge.

99

16 Comments

A developer in San Francisco selects MiniMax as her inference backend. The request crosses the Pacific, gets processed in a Chinese data center, returns in a second. She pays $0.30 per million tokens. Claude Opus would cost $5.00 for the same volume. No customs declaration. No tariffs. No entry in any trade database. Chinese tech media calls this Token 出海: Token Export. This month’s OpenRouter data showed it crossing a threshold: Chinese models surpassed American ones in total token consumption for the first time. This did not happen overnight. I spent February tracing the structural forces behind it: efficiency architectures, energy cost gaps, and a competitive intensity that pushes pricing below rational margins. Nvidia’s record quarter and falling stock price are part of the same story. China’s Compute Bet explains the system. https://lnkd.in/ghEBwk-s The Export That Tariffs Can’t Touch explains what happens when that system meets the global market. https://lnkd.in/gh_ajmmZ

3

Buried in the Claude Haiku 4.5 system card was a really interesting nugget that most people missed, but it could have a huge impact for anyone building with LLMs “For Claude Haiku 4.5, we trained the model to be explicitly context-aware, with precise information about how much of the context window has been used. This has two effects: the model learns when and how to wrap up its answer when the limit is approaching, and the model learns to continue reasoning more persistently when the limit is further away." For years, context awareness and management have been handled by developers and it's one of the hardest parts of building agents. Teams built systems to count tokens, monitor context usage, and trigger downstream functions like summarization or condensation. With Claude 4.5 Haiku, context awareness is now trained into the model itself. It reminds me of what happened with Chain of Thought. Something that started as a prompt engineering hack but eventually became part of model training (reasoning models). There’s still plenty of work to do around context management, but if this is the direction things are heading, the next major leap in model capability may already be taking shape: models that can manage their own context intelligently. S/O to Simon Willison who originally highlighted this on his blog (linked below)

22

3 Comments

Super exciting to see reports this morning that Waymo is in advanced discussions to raise $10–15B at a valuation near or above $100B, with Alphabet expected to anchor the round. That would be a major step up from Waymo’s prior ~$45B valuation in 2024 and a clear step change in investor confidence. Notably, Waymo has not raised this year, despite the broader AI funding frenzy. The size and structure of the round also make this feel like a step toward an eventual spinout or IPO. This matters because capital supercharges growth. Waymo is past the “does this work?” phase and firmly in “how fast can we scale?” mode. The company has already logged ~14M rides this year and expects to reach ~1M rides per week by late 2026 as it expands into markets like Miami, Dallas, and Philadelphia, including expanding onto freeways. At the same time, competition and consumer expectations are accelerating. In markets where Waymo operates, fully driverless rides are becoming normal rather than novel. Once people can reliably hail a car with no one in the driver’s seat, tolerance for “pilot programs” and distant timelines drops quickly. The race is no longer about whether robotaxis work. It is about who can scale them first, safely, and everywhere — a win for consumers who get cheaper, safer, and more reliable rides.

25

Why does the SF Bay dominate the tech landscape? There's a network effect, like a gravity well that creates compounding benefits from those that are within the region. Here's one of my slides from my TEDAI San Francisco speech on why community and culture play a key role in the AI ecosystem development in our region.

60

7 Comments

AnySignal just dropped a Series A that feels less like a broadcast to anyone building in space, defense, or the increasingly loud intersection of both. A $24M raise led by Upfront Ventures is not casual capital. It is a signal that the market sees what John Malsbury, Jeffrey R Osborne, and Ricardo Medina have been assembling since 2022: a vertically integrated RF powerhouse that can pull mission-critical data out of the most unforgiving environments and make it look routine. When a team stacks SpaceX DSP leadership, Kepler strategy firepower, and #cloudinfrastructure mastery under one roof, you stop calling it a startup and start calling it a force multiplier. What catches my attention is how quickly AnySignal turned credibility into momentum. 4 missions already in orbit. 12+ booked for 2026 across #LEO, #GEO, and #lunar routes. A customer backlog thick enough to push their El Segundo production line past its limit, forcing them to scale from ~100 radios a year to a target of 1,000–2,000 once they move into a larger LA facility. That kind of demand is not a marketing win. It is a physics problem, and AnySignal is solving it with the kind of discipline usually reserved for legacy aerospace giants, except faster and with far fewer meetings. The Iris platform is where the story gets even more interesting. It blends #comms, #RFsensing, #radar, #navigation, and adaptive DSP into a single system that behaves less like a device and more like a mission partner. The AnyLink network layers on top, turning cross-network chaos into unified, intelligent #connectivity. In a sector full of promises about "interoperability," AnySignal is one of the rare teams actually delivering it. When your hardware and software can traverse #satellites, ground stations, and edge systems without throwing a tantrum, customers notice. What makes this raise feel inevitable is the way the founders' backgrounds slot together. John Malsbury brings years of DSP and FPGA leadership across Dragon, Starship, Starlink, and national security programs. Jeffrey Osborne adds the strategic instincts of someone who has already built a #spacecomms company from zero. Ricardo Medina ensures the cloud backbone does not buckle when missions scale. Add investors like BlueYard Capital, First In Ventures, LLC, Also Capital, Acequia Capital, Balerion Space Ventures, and Caffeinated Capital, and you get a syndicate that knows real dual-use infrastructure when they see it. If you are building #spacecraft, #nationalsecurity systems, #climatesensing missions, or autonomous platforms that cannot afford static when the stakes spike, this company is worth more than a follow. AnySignal is not getting louder by accident. It is getting louder because the market keeps leaning in. #Startups #StartupFunding #VentureCapital #SeriesA #Aerospace #Defense #DefenseTech #Space #SpaceTech #DeepTech #Data #DataDriven #Infrastructure #Cloud #CloudComputing #Technology #Innovation #TechEcosystem #StartupEcosystem

3

Generation is solved. Output is now cheap. The bottleneck is knowing whether that output is safe to act on. That's the verification problem. And it's almost entirely solved. The three things we're not admitting: -The hard part is upstream of any tool. Before you can verify behavior, you need a definition of correct. Ask engineering, product, legal what correct means for your agentic system. You will get answers that don't reconcile. The role that owns that definition doesn't exist on most org charts. Until it does, you're not verifying anything. -Evals degrade through use. The moment you optimize your agent against an eval suite, you've contaminated it. The agent adapts. Performance improves. The signal dies. Same thing that hollowed out every foundation model benchmark. Verification is a race you run permanently. -The deployment team and the verification team have opposite incentives. One wants the agent in production. One wants to find problems. Left unmanaged this resolves one way: deployment pressure wins. Verification becomes a gate to route around. The generation problem got solved faster than anyone expected. But the verification problem is organizational, adversarial, and almost entirely unbuilt. Interestingly, cybersecurity solved this exact problem a decade ago. You cannot enumerate all attacks in advance. Static defenses work until they don't. So security stopped trying to enumerate and started building adversarial infrastructure that finds failures continuously. Red teams. Penetration testing. Behavioral anomaly detection. The shift from checking against known failures to searching for unknown ones. AI verification is ten years behind security. The playbook exists. The industry just hasn't stolen it yet.

124

46 Comments

https://lnkd.in/g6JBTDux “Steam, Steel, and Infinite Minds.” Ivan Zhao’s latest piece stopped me in my tracks. It made one thing painfully clear: we are entering the first era where human progress is no longer constrained by the scale of human cognition. Just as steam and steel propelled us from agrarian labor into industrial civilization, AI is introducing a new kind of material into our world: Infinite Minds. This isn’t a faster tool. It’s not smarter software. It’s a fundamentally new production factor, one that can multiply human thinking exponentially. A few realities we can no longer ignore: 1. AI is not an efficiency tool. It’s a cognitive amplifier. When thinking itself can be replicated, parallelized, and run continuously, team size no longer defines intellectual capacity. The real advantage isn’t having AI, it’s knowing how to collaborate with infinite cognition. 2. The office was never the final form of work. Human labor moved from factories to desks. AI may now dismantle the idea that humans must be the sole cognitive center of work. Knowledge workers won’t disappear, but roles built purely on individual brainpower will be fundamentally reshaped. 3. Tools are becoming the factories of the future. As Ivan points out through Notion, we are no longer building “software.” We are constructing cognitive infrastructure for the next civilization. Ignoring AI isn’t cautious, it’s choosing to stand against the direction of history. At Genomii, this question confronts us every day: When AI stops being “advice-giving software” and becomes something that can continuously understand, reason, execute, and stay with humans through long-term goals, where does true human value belong? I’m increasingly convinced of this: The most important skill of the future isn’t knowing how to use AI, but knowing how to form long-term collaboration with it. If steel built empires, what kind of civilization will infinite minds create? This isn’t science fiction. It’s already unfolding. And I choose to see it as an opportunity, an opportunity to redefine the human role itself. What do you think? #AI #InfiniteMinds #FutureOfWork #Genomii #Leadership #AGI #notion

3