I’m not sure I agree with this but maybe I just lack self awareness?
There are large portions of my codebases that are essentially extremely verbose grunt work. My UI stack, IaC YAML, thin CRUD routes, etc.
I know what the code is supposed to look like when it’s done being written, but it’s going to take me for freaking ever to type it all out.
I can just few shot it now in an hour. Plan -> feedback loop -> build -> review loop.
Does it try to do weird stuff? Yeah. And then I’m just like “that’s weird, no, the components should be broken up like XYZ” and then it’s not weird anymore. Occasionally (1% of the time) I just do a quick refactor myself instead of trying to tell the agent harness what to do.
I can get something fairly close to the ballpark of what I would have done but in like single digit percentage of the time.
And the result is that I can spit out a bunch of purpose built tools (personal tools, internal tools for teams, etc.) that I never would have been able to justify building otherwise.
Anyone can write software, you can't stop them. What we can gatekeep is the building, distribution, installation, and running of software that affects critical systems, like one of the most popular OSes.
The XZ backdoor affected millions of computers, with the potential to effect hundreds of millions of computers, many of which had the capacity to affect billions of people. From one completely unregulated software library.
I've never used breakpoint debugging, was always a printf debugger. And now an agent can do that loop for me.
Prompt is usually something along the lines of:
>I would expect the behavior of this to be [X] - instead I'm observing [Y]
And the agent will form hypothesis, place printf statements, compile, and scrape logs on loop - each loop ruling out hypothesis or narrowing down what portion of the code is responsible for the unexpected behavior.
It has been able to pin-point the exact line(s) of code responsible every time I've reached for it so far.
> If your data and workloads live on standard Kubernetes with Postgres, object storage and the usual Prometheus/Grafana/Loki, then no single provider (EU or otherwise) actually has you over a barrel.
I’ve heard this refrain most of my career - and believed it at one point. But the bytes themselves are heavy and hard to move.
Multiple projects I’ve worked on in my career would have had to _physically move_ the data (plane, train, semi) if they wanted to migrate - there was no reasonable way to get the data out of the datacenter over network on any reasonable timeframe.
> This technology is fundamentally different from the “rivulet (shark skin) process,” which is known as a typical aerodynamic drag reduction technology. The rivulet process mimics the fine longitudinal grooves in shark skin, and by carving grooves approximately 0.1 mm wide along the direction of airflow, it aligns the vortices that occur near the wall surface of turbulent airflow areas. DMR, on the other hand, delays the switch from laminar to turbulent flow by means of random and minute irregularities. The flow zones it affects and the mechanisms it employs are based on completely different concepts.
I think I noticed it reach “end game.” The snake reaches a point where, if it gets any longer, it is out of squares and hits its own tail. So it finds the route through the squares that it can infinitely loop, never eats the ball, and score starts dropping and goes negative.
There are large portions of my codebases that are essentially extremely verbose grunt work. My UI stack, IaC YAML, thin CRUD routes, etc.
I know what the code is supposed to look like when it’s done being written, but it’s going to take me for freaking ever to type it all out.
I can just few shot it now in an hour. Plan -> feedback loop -> build -> review loop.
Does it try to do weird stuff? Yeah. And then I’m just like “that’s weird, no, the components should be broken up like XYZ” and then it’s not weird anymore. Occasionally (1% of the time) I just do a quick refactor myself instead of trying to tell the agent harness what to do.
I can get something fairly close to the ballpark of what I would have done but in like single digit percentage of the time.
And the result is that I can spit out a bunch of purpose built tools (personal tools, internal tools for teams, etc.) that I never would have been able to justify building otherwise.
reply