Alex Hernandez

when i said fully agentic hardware i meant FULLY

The next AI database isn’t another database. it’s a layer over the one you already have. Today we're launching a virtual ai search layer you can deploy over ANY Postgres database. no migrations. no vector DB. no Neo4j. no sync. no bs. One database. Every way to search it. This is virtual search that feels like extended context. try it free → polygres.com @daltonmeon @damienhe @polygres

how many times a week do u go from “we are so cooked” to “we are so back” ?

WE'RE OPENING PRE-ORDERS STARTING TODAY!!! First batch: 100 devices, pricing: $180 (25% off!) Next batch: 500 Devices pricing: $210 (12.5% off)! FULLY REFUNDABLE BEFORE SHIPPING. RESERVE YOUR SPOT NOW! LINK IN THE REPLIES You get to keep the device fully, so don't wait until your competitor buys it up! Seems like everyone's wanting the ACTUAL DEVICE more than the remote shell access, and it's why we made a pre-order link right after launch. We shut down the “deposit for remote shell access” which we can't deliver at the moment due to the software corruption (it’s stuck in maskrom instead of booting)… The next iteration will be far more polished than it is right now, and we'll make sure it doesn't shut off randomly. We cannot fully guarantee the exact lead time, but here's the rough estimate we have for the first batch: Final design changes: 4-8 days JLC pcb manufacturing: 3-6 days JLC pcb assembly: 2-6 days Camera mount, calibration, case assembly, packaging: 2-3 days Shipping: 4-6 days Estimated min: 4 + 3 + 2 + 2 + 4 days = 15 days Estimated max: 8 + 6 + 6 + 3 + 6 days = 29 days (Customs are not included in the estimate)

Asked what he would build if he started over, Patrick Collison did not name an AI lab. He said he'd go for "the janky 1990s software" running whole industries nobody in the Valley will touch. 3 examples of this: 1) Vertical SaaS

I found a great example of how experimental data is crucial for AI in Science specifically for Material Science explained by @josephfkrause(Cofounder @RadicalAI ) : Let's take an example of materials, you have to first make it, test it, characterize it, and see if it can go into real application(can be used in industries ) -> so for that you can build a close loop system(Self Driving Lab) -> RUN those experiments -> CAPTURE the data -> FEED that information back to that AI scientist so that they can learn and predict the materials relevant to industry. what makes a material real is in the latter part of discovery phases, specifically at characterization and synthesis phase. so we have to ask questions like: what did we make? does it have cool properties in the lab ? in domains like structural metals/alloys -> performance of alloys is governed by the processing -> how do you manufacture, what techniques you using in post-processing and manufacturing. It can push or change the performance. You can generate a new composition using AI. You have a new discovery but can it be actually used in application space you are getting into? can you actually make it? can you scale It(can be used by actual end customer)? the last 2 questions can't be solved using AI right now so a model can't figure out a way through the qualification pipeline for a new alloy for a jet turbine -> you have to do experiments to do that. experimental data == ground truth we know what we want to make. are we actually making those things, and do we actually have the properties we care about , can we actually push them to industry. the latter questions are the hardest questions in materials. another problem in this industry is "long timelines" typically 15-30 years, why is it so fragmented in materials today is because: academia focuses on -> discovery small companies (supported by DoE, Department of War, other gov programs like NSF) on -> testing late stage/big companies on -> optimizing their current systems to get margins. Result of this fragmentation -> the data never gets shared and the connection from discovery to manufacturing is lost in that process. that connection needs to be established and that is the opportunity for AI and autonomy in material to start linking these two making a close loop systems.