Lenny Pruss

Our OCR Models are really good at extracting text and other visual information from any image into structured data! One of our partner energy companies recently tried using it to extract structured data from plates on transformers and it worked surprisingly well. From the very…

It begins...

what many still don’t seem to grasp is something U.S. military strategist @SpencerGuard has emphasized: the Muslim Brotherhood and its ideological offspring have pioneered a new form of warfare, one in which the spectacle of civilian suffering is not an unintended consequence,…

Need to pull tables, metadata, signatures? Tensorlake extracts structured fields, all you have to do is provide the schema.

If only they could rewind time... pcgamer.com/software/ai/i-…

TBH, I'd put my money on something new. Something strongly typed, that isn't verbose (context window), with proof-like qualities that something like Rust has. IMO, there's potential for innovation here. Similar to my belief that LLM should speak substrait, not SQL.

Excited to share a fun @gensx_inc / @GroqInc demo! 📝Draft Pad 📝 – A fully voice controlled canvas experience where you can iterate on text fast without the back and forth chat. And you can compare multiple model outputs concurrently which is super useful too

Tired: AI Agents Wired: Authorization for AI Agents In case you missed it: Here's how you can prevent Agentic AI from accessing unauthorized data: (demo & batteries included) youtube.com/watch?v=UjKa5T…

IronBeetle⚡️ Ep. 075, July 3rd Unsurprisingly, attempting to explain our deterministic performance testing was foiled by the need to debug a performance issues live youtu.be/UwowBtfiN1I

Agentic apps represent a radically new architecture that promises to strain existing infra, in particular the database. @StatelyCloud explains... stately.cloud/blog/the-futur…

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

Some practical examples: - Galois has a lot of tooling for formal verification and did a project for Amazon s2n: galois.com/project/amazon… - I think Wireguard has formal verification in its CI/CD system using Tamarin: wireguard.com/formal-verific…