Sponsored by
Welcome, AI enthusiasts
Peter Thiel just led a $140M round into Panthalassa, a startup building floating AI compute nodes powered by ocean waves. With grids jammed and permits slow, Silicon Valley is starting to look offshore for the next wave of data centers. Let's dive in!
In today’s insights:
Peter Thiel's biggest AI bet is headed out to Sea
Coinbase Cuts 14% jobs as CEO Armstrong Goes AI-First
Grok Got Tricked Into Sending $200K in Crypto Through Morse Code
Read time: 4 minutes
LATEST DEVELOPMENTS
PETER THIEL
🌊 Peter Thiel's biggest AI bet is headed out to Sea
Source: Panthalassa
Evolving AI: Panthalassa raised $140M to build floating, sea wave-powered AI nodes that bypass the grid entirely.
Key Points:
Series B led by Peter Thiel, with John Doerr, Marc Benioff's TIME Ventures, and Founders Fund joining in.
85-meter steel nodes generate power at sea, run AI inference onboard, and transmit results by satellite.
Ocean-3 pilot deploys in the northern Pacific in 2026, commercial fleet planned for 2027.
Details:
Ocean-3 is the next step in that 2026 pilot, an 85-meter steel node stamped from plate steel at coastal factories and sent into deep-ocean wave zones where it generates clean power around the clock. Rather than sending electrons back to land, the node runs AI chips onboard and beams inference tokens to customers via low-Earth-orbit satellites. Total funding now sits at roughly $210M completing the pilot factory near Portland and building on the Ocean-1 and Ocean-2 prototypes tested in 2021 and 2024.
Why It Matters:
A node that powers itself starts to look less eccentric when you see what land-based builds are running into. Roughly half of global data center projects slated for 2026 are delayed by power constraints, and Morgan Stanley sees a 49 GW shortfall in the US alone by 2028. Microsoft is signing direct gas deals with Chevron to keep its projects on schedule. Panthalassa is making a bet that open water is faster than open permits.
Benchmark against 2,000+ private B2B SaaS and AI companies
Is your growth in-line with your peers in B2B SaaS & AI?
Benchmark yourself against actual billings data for Maxio’s 2000+ global customers, alongside firsthand company perspectives to understand how growth varied by company size, business model, and strategic focus.
Key takeaways from the report:
Average growth across 2,000 companies
Growth by revenue band
AI-led vs AI-enhanced. Who performed better?
Evolving AI: Coinbase is laying off around 693 employees in an AI-driven restructuring.
Key Points:
Armstrong wants 50% of internal code written by AI, making Copilot and Cursor mandatory.
The org chart is being flattened to five layers max, with one-person "AI-native pods" replacing traditional teams.
Around $50M to $60M in restructuring charges expected in Q2.
Details:
Coinbase announced the cuts as a pivot rather than a retreat. Armstrong's pitch is that engineers now ship in days what full teams used to take weeks. Pure manager roles are being scrapped, leaders have to stay on as active individual contributors, and the new AI-native pods fold engineering, design, and product into a handful of people, sometimes just one. As affected US staff get 16 weeks of severance plus two weeks per year of service, and Wall Street liked the discipline enough to push shares up over 4%.
Why It Matters:
The market reaction is what makes this interesting. Coinbase posted a $667M loss last quarter and Bitcoin is down a third from its October peak, so analysts are already arguing the AI story is doing cover for plain old crypto winter pain. Yet investors clearly preferred the AI version. The cyclical pressure is real, the AI productivity gains are real, and somewhere in between sits a much more flattering narrative for everyone holding the stock.
Stop re-prompting. Say it right the first time.
Voice-first prompts preserve the nuance you cut when typing. Speak once, paste into any AI tool, get results that don't need a follow-up. 89% of messages sent with zero edits.
Source: Dexerto
Evolving AI: An X user used Morse code to bypass Grok's safeguards and trigger a 3B token transfer.
Key Points:
The exploit chained two AI agents, with Grok acting as a translator and Bankrbot acting on the translation.
An NFT gifted earlier had quietly expanded what Grok was allowed to authorize on chain.
Roughly $200,000 in tokens moved before the attacker sold and walked away.
Details:
The attacker first sent a Bankr Club Membership NFT to Grok's wallet, which expanded its permissions inside the Bankr system. They then posted a Morse code message on X and asked Grok to translate it for Bankrbot. The decoded text told Bankrbot to send 3 billion DRB tokens to a specific address. Bankrbot read the output as a real instruction and executed the transfer. The attacker sold the tokens and deleted their X account. The funds were eventually returned, but the path was already proven.
Why It Matters:
The interesting shift here is, that exploits no longer need a single vulnerable agent. They need two agents who talk to each other. As multi-agent setups spread across Web3 and beyond, the weak point keeps moving from the model to the connection between them. A prompt can guide behavior, but as one researcher put it, a prompt is not a control. That gap is where the next wave of incidents will live.
👀 Click on the image you think is real
QUICK HITS
🦾 Meta is building an OpenClaw-style AI agent aimed at everyday users.
⚡ OpenAI rolls out GPT-5.5 Instant as ChatGPT's new default model.
💶 SAP commits $1.16B to German AI startup Prior Labs while greenlighting Nvidia's NemoClaw.
🛍️ Etsy launches a ChatGPT native app for conversational shopping.
📈 Trending AI Tools
*partner link