ctstoner · HackerTrans

10 tháng trướcdiscuss

The difference between Zenode and something like Perplexity is that Zenode is built specifically for electronics, so it's trained specifically for searching electronics, and the data is much more detailed for electronics. Perplexity has to be good at everything; Zenode is specialized, so it just needs to be the best at working with electronics!

ctstoner

10 tháng trướcdiscuss

We typically take the 'nominal' value in the datasheet's electrical specs table for parametric specs. If the manufacturer is trying to be sneaky by providing the value with perfect test conditions to improve the numbers, we still give the value because there's no other way to give a credible answer.

However, we saved all the test conditions and similar notes. Maybe we could display the conditions with the specs in the table? That way, any outlier manufacturers would jump out right away.

ctstoner

10 tháng trướcdiscuss

You should be able to drop the specs you need into the zenode search and get some options; there's always so many choices for buck converters!

Are you looking at a multiphase design? I think I've seen a few chips that control multiphase converters: You'd get a much higher current design with just one IC (and a few extra inductors).

ctstoner

10 tháng trướcdiscuss

Agreed - trust is the key! That's why we've built in sources with links to the exact location in the datasheet and part documents where the AI found an answer. We're working hard to make sure you can trust its answers, but we know most engineers 'trust but verify'. A (transparent) confidence score is a great idea to improve trust in the answer and sources.

To close the gap, we've built our own Q/A datasets and are training custom AIs how to search and read a datasheet (like a new engineer needs to learn early on). We're concentrating on teaching the AI how to identify key information vs noise as it relates to electrical engineering (differences like 'Voltage' in the Absolute Max vs Recommended section) and where information is likely to be found in a datasheet or app note.

ctstoner

10 tháng trướcdiscuss

Who we are We’re electrical engineers who have lived this pain firsthand for a decade; Brandon (OP) has designed a dozen boards, while I have built over 250 through production. Years ago we spun up a brand-new product line in two weeks to help salvage a $150M deal. That meant 18-hour days spent picking parts, doing CAD, paying $$$$$ for three-day prototypes and praying everything worked (one did blow up when we accidentally shorted 3 kV into a metal screw on the desk, we nearly died cause of a sleepy brainfart :dizzy_face:). Even from that working version, it took an additional four months of redesigns to get through certifications and into production. When ChatGPT shocked the world in ’22, we realized that the initial superpower of AI in electronics isn’t in generating designs (too risky — one mistake kills the board), but in digesting the endless unstructured documentation that plagues (benefits? :thinking_face:) our industry.

comments