Write a decent comment, for one. It struggles with some of the firmware stuff I do, though that's more due to a lack of a proper automated feedback loop.

Yeah, it's generally the harness around the model that makes it useful.

Many of my tests after a certain point were open book, or at least explicitly allowed a cheat sheet.

It wasn't going to help you if you didn't know the material, though writing the cheat sheet in the first place is a pretty good form of studying.

Hand writing assembly produces more efficient and concise code, at the cost of developer time and required expertise. It was true for a long time, now not so much.

Is this sarcasm? I assume most studios are integrating AI into their workflows, but I still haven't seen a single vibecoded game that looks interesting.

You can get a pretty good 1080p60 video at only 6mbps. Depends on how well the sender is encoding I guess, but it shouldn't be too much more than that.

I think the drones don't use a complex encoder because latency.

Some of the more basic ones are a very rudimentary NTSC analog video in to H264 encoder board that is tiny in weight and power consumption. They're using the same NTSC analog video-out forward flight cameras as are used in pure analog FPV. It's also not doing really 'complex' H264 encoding. Very similar to what you would get for the H264 encoder in a cheap $60 IP (RTSP) security camera.

There's more advanced ones that take video input from a MIPI digital video interface or native HDMI input.

Not just because latency but also because of cost, size, weight, power.

Rotation data isn't included in the file, or it just doesn't display with the correct orientation?

Sometimes Google photos shows the wrong orientation for my pixel photos/videos but it's just a display issue.

Is the core of your argument that the industrial revolution and other such changes should have been aborted due to their downstream negative affects?

They also lead to great advancements in quality of life, and the capability of sustaining much more human life. We can't predict the long term outcome of new technologies, so the best we can do is blindly forge ahead and try to mitigate the obvious short term problems.

Yes, industrial revolution should have been managed differently and yes the world would be better if it was.

Easy to say "should".

This is difficult to read, maybe just link to a gist?

That's because it's copy pasted from a coding agent.

It's difficult to read because it doesn't have line breaks.

Around 20s saved by eating on those \n\n.

It looks at least partially hand edited to me, although it's getting pretty difficult to tell with Astra...

Thank you! Here is the short version:

  Kimi K3, 2.78T parameters, ~1.45 TB of MXFP4 experts streamed from four SSDs on an M5 Max / 128 GB. 1.00 tok/s steady over 512 tokens, 1.13 over 128, ~6.3 min to first token on a 512-token prompt. Output token-identical drafter on/off on a given drive layout; the int8 trunk is non-weight-exact per upstream.
  
  The useful bits: one drive gives ≈52% of four, two ≈73%, three ≈90%; and prefill is slow because of ~9 TB of reads for a 1.4 TB model — a scheduling bug with a planned fix.
  
  README with per-run logs: github.com/argonautlabsai/deltafin — a fork of gavamedia/deltafin, who built the engine.

Can't rule out privilege without knowing a bit more background. Most people work hard, it increases the odds of success but isn't necessary.

Sol and Opus are pretty good with the right prompting and feedback framework.

Or, my standards are lower. It's sometimes hard to tell in these discussions whether people are talking about getting production quality results, or stuff that's good enough for a one off blog post.