They can be expensive to maintain

Love to see it in action

If there was low visibility and you are trying to get someone down he trail that is unprepared or just plain freaking out, you could send a drone with a pair of AR glasses and improve their ability to get down the trail. I would love to have something like this for snow sports to show the best trail to follow.

Same thing happens with humans but you don't get the warning not to fabricate. Bodge wires ftw.

I like this idea, keep going.

5 minutes, $0.52 in API fees and $5 each for fully assembled boards. Free and open source WholeBoard designed, laid out, checked inventory, cost reduced the BOM, simulated and documented a simple board that flashes 2 LEDs. It runs on your PC with local or cloud AI. Got an idea for a new board? What's stopping you downloading and trying WholeBoard?

I hope this is a real demo and not simulated. This is the sort of real work tech that I want to see brought into existence

That is cool

Yeah I am really intrigued by how quiet Xai are abotu the next version of Grok. 4.7 was meant to be a big jump on engineering tasks but it wasn't. Are they still fighting for the frontier?

I have since made an open source free workflow tool that uses AI that will help you make your own board if interested: github.com/DingoOz/WholeBoar…

I made a free open source tool for this now: github.com/DingoOz/WholeBoar…

I feel like some of the audience might not know what building a dataset in a domain means without added an example early on, something like "Need to turn your 20 years of hydraulic calculations into a personal AI agent?....

Yes I am just playing around to see where models can do things. I am really looking forward to a local model that has the same spatial awareness as Opus 5.5

Same weekend, three different models, four games. Qwen4-27B made the Atari-era style brawler. Qwen3.8-Flash-Next made the neon Tetris. Claude Opus 5.5 made the rally game and the 3D mech sim, and those two look PS2 level to me. The two Qwens run locally. Which one would you want to see build the next game?

Extraordinary claims require a similar level of evidence

It should handle Ollama models just fine. If it doesn't let me know the issue and I can fix it.

pedal to the metal!

And the fact it is a preview for Qwen4 is such a good omen!

~100 days ago Grok 4.5 came out and I thought it was a massive breakthrough. Today I'm running Qwen3.8-Flash-Next at Q6 on 10 year old hardware at ~40 tok/s, and on the Artificial Analysis Intelligence Index it scores 40 vs Grok 4.5 (high) at 39. Same chart attached. This is moving so fast.