It’s a really interesting project, worth checking out the video where he merged it with a companion computer to do computer vision tasks and inject controls straight into the flight controller https://m.youtube.com/watch?v=uaY2G5Kbj_g
True, certainly something you could do with Tailscale+tmux but if this makes that a lot easier then it could end up being popular, increasingly the people who are pushing the most money into the likes of Anthropic are not rhe people focusing too much on the amount of money being put towards achieving their goal, rather the progress(tbh more realistic to say perceived progress) towards that goal in the shortest amount of time.
Fair enough but the examples shown could surely just display some pre-cooked examples to give a demo of how rhe product works with no real cost to you or barrier to potential users.
Confused me at first as when I saw mention of local + the single file thing in the GitHub I assumed they were going to have llamafile bundled and went looking through to see what model they were using by default.
100%, trying a bit of an experiment like this(similar in that I mostly just care about playing around with different agents, techniques etc.) it has built out literally hundreds of tests. Dozens of which were almost pointless as it decided to mock apis. When the number of failed tests exceeded 40 it just started disabling tests.
To be fair, many human developers are fond of pointless tests that mock everything to the extent that no real code is actually exercised. At least the tests are fast though.
Citing the absolute worst practices from terrible developers as a way to exonerate or legitimize LLM code production issues is something we need to stop doing in my opinion. I would not excuse or expect a day one junior on my team that wrote pointless tests or worse yet removed tests to get the CI to pass.
If LLMs do this it should be seen as an issue and should not be overlooked with “people do it too…”. Professional developers do not do this. If we’re going to use Ai for creating production code we need to be honest about its deficiencies.
And most of the big players now have some kind of browser or bowser agent that they could just leverage to gather training data from locked down sources.
^This 100%. Junior SWE here. Agentic coding has kinda felt like a promotion for me. I code less by hand and spend more time on the actual engineering side of things. There’s hype in both directions though. I don’t AI is replacing me anytime soon(fingers crossed), but it’s already way more useful than the skeptics give it credit for. Like most things the truth’s somewhere in the middle.
In terms of training your own models, is there enough COBOL available for training or are you going to have to convince your customers to let you train on their data (do you think banks would push back against that?)
reply