🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

We surveyed hundreds of engineers building in AI about everything from which models they’re using to whether they’re using a dedicated vector database. And, of course, if they think everyone will have AI girlfriends by 2030. Some highlights 🧵:

Audio wave incoming

Knowledge maxxinggg.

If you already know all the answers then I won’t bore you with the facts. For those interested in the facts, behold Barr’s latest AI Engineering survey 📝

Super cool insights on AI usage!

It can be so tough to understand if you’re building the right thing when every company claims to be running multi-agent systems with tool use and reasoning capabilities in prod. Barr’s survey cuts through the cr@p to share insight into where we actually are.

really awesome survey from Barr!! highly recommend people read to see the state of ai eng

🥖 Breadcrumbs are out. Full loaves are in. In other words: bitdrift now supports first-party crash reporting! Check it out here → buff.ly/vjJNlII

I had sleepless nights for months before our first production release, wrestling with the problem: How to guarantee safe upgrades? Watch Federico Massimiliano Lorenzi's talk about the hardest problem in TigerBeetle: youtube.com/watch?v=P9nLS2…

We’re excited to introduce Chai-2, a major breakthrough in molecular design. Chai-2 enables zero-shot antibody discovery in a 24-well plate, exceeding previous SOTA by >100x. Thread👇

New survey data from Amplify Partners Barr Yaron shows 17% of respondents change models weekly and 10% of respondents change their prompts daily! With such frequent changes, evals at a specific point in time are no longer moats. The tooling to runs evals are..

Quoted Elliot Hershberg on the era of digital biology; the Amplify Bio Manifesto: amplifypartners.com/blog-posts/amp… Amplify Partners

Every engineer is familiar with the joys of writing unit tests. But less known is formal (or program) verification: foolproof, mathematical ways of guaranteeing code outputs. And as more code writing shifts to agents, I believe it will become critically important for LLMs.

Formal verificiation 👀

Lenny Pruss Formal verification (theorem proving and model checking) is often considered programming's holy grail. In my view, research remains impractical (see halting problem). Humans often fail to test boundary conditions and nondeterministic results, which is the sweet spot for LLMs

The holy trinity of serverless GPU has been achieved internally at Modal: ✅ Custom filesystem optimized for container cold start ✅ CPU snapshot+restore ✅ GPU snapshot+restore Last one is working internally but not released yet – stay tuned!

There’s an open question around which programming language is going to become the standard for AI agents generating code. Weakly typed ones like Python + JS have leapt ahead because they’re easier for our meat computers to grok…but what happens when it’s not humans writing them?

We talked to Elliot Hershberg (Partner Amplify Partners) about the challenges in life sciences technology. "There has been a general sentiment that you can make a huge amount of progress with new data and new technology in life sciences." "It actually turns out that it's really

Our very own Elliot Hershberg on TBPN: