I think anthropic is trying to protect their moat by preventing other labs to use Fable 5 aka Mythos for doing training infrastructure and chip design. A new anthropic chip in progress?
Overall bullish for AI infra companies as big labs are worried about the moat here.
Never a boring day in AI infra,
> $GOOGL casually raising $80bn via selling its own shares for building data centres. Looks like the private markets have already been tapped by recent Anthropic raise!
> Also, Anthropic files for IPO, the same company raising at $185bn last September 🤯
Again, AI supply chain is fully nuts.
startups which positioned themselves as “shift left” for infra/dev tools and created a feedback loop on DevX are going to be great beneficiary of the AI wave. “Shift left” traditionally had two essential ingredients which amplifies agents adoption:
- Putting key SDLC components in hands of devs; infra provisioning & management, testing etc.
- Removing software team silos. No individual testing/infra teams and no scattered context.
Agents get complete and rich context and delightful cli interfaces to get their job done. @harnessengr is a great example of shift left execution and philosophy.
Finally got the secure and fully local hermes agent setup working after a few hours of dabbling,
- parallel agents spawning using multiple sandboxes, load balanced with hot pools
- Fully local LLM Gemma-4-31b optimised for faster prefill and concurrent sessions
- Everything behind an API gateway, fully auditable
- Always on and working for you, lets gooo..
AI supply chain is nuts.
Just wrapped a call with a founder friend selling to ai labs and enterprises, they are doing way over 100M ARR in less than a year. Raised less than 10M.
@mlejva@e2b Totally relate to this. Have been trying to onboard a few founder friends around what next generation of agentic infrastructure is gonna look since last October. Sandboxes adoption is just a starting point and no brainer.
We've raised $65 billion in Series H funding at a $965 billion post-money valuation, led by @AltimeterCap, Dragoneer, @Greenoaks, and @sequoia.
This investment will help us advance our research and expand our capacity to meet growing demand for Claude.
The mental model I use,
- Sandboxes are cloud computers where you can long in and log off any time. You pay only when you use.
- Agents are on demand employees working in sandboxes before logging off and waiting for your input. Recent patterns allow scheduling like crons as well.
For a startup registered in SF and have VCs and clients in US, why do paid partnership and profile pin on the Indian Tech & Infra page?
Sums up the indian AI ecosystem. Good luck to the founders btw.
🚨 AI’s next infrastructure race has started.
Trajectory launched today with $15M from Conviction, Bessemer, Radical Ventures, Jeff Dean, Fei-Fei Li and more to build the platform for continual learning.
The company is focused on turning product usage into training signal, allowing agentic systems to improve continuously after deployment.
As models commoditize, the systems that learn fastest may become the real moat.