The Claude models are the best by far for coding assistance in my experience. Apparently a lot of other people think so too because Anthropic is getting swamped. They are having to ration out their compute resources and in some cases have raised their fees to 2-3 times more than the lesser competition charges. I'm finding that in order to keep costs down I'm having to use 2nd-tier models for simpler work and revert to Claude for the heavy lifting. A hassle.

Clearly the demand is there. At this point I expect Anthropic is revenue-limited by their infrastructure availability so it makes sense that they recruit the big players to help beef it up.

>> Was giving you an out

That's an obvious lie, and I stand by my statement.

>> you showed nothing

Another obvious lie. I merely showed something you didn't like.

Lol, now you're accusing me of not writing my own posts here? Hilarious!

>> The competition is easily keeping up with Claude Code

Show evidence. Like I did.

"Claude Code has gone from zero to be the #1 tool in only eight months."

https://newsletter.pragmaticen...

Looks like OpenCode is #7.

There are some significant advantages over Python. JavaScript on Node.js (or Bun in this case) is inherently event-driven. TypeScript gives strong typing on top of that. And apparently Bun can package up an app into a standalone executable, also unlike Python. I'm going to have a look at Bun, seems powerful.

My sense of it is that this code is not the LLM itself, it is the infrastructure and interface layer between the user and the LLM. But even so, there's a lot of expensive work that has been exposed. At the very least it will give the competition some ideas.

>> That's because I know better.

You obviously don't know a damn thing, lol.

>> Rocky's echolocation can even work through walls

Yeah, his echolocation can work through impervious xenonite walls and then through a completely different atmosphere/pressure on the other side. Give me a break!

>> Executive direction that the legacy codebase must be 'documented' fully

I'll assume you don't work in the profession, the execs usually don't care or even know. If they had wanted documentation they would have emphasized it when the code was being written, or at least shortly afterwards. Software rots over time and if its a valuable internal application someone will eventually have to fix it. I've been paid very good money at times to do that job.

Sometimes the first task is to determine where the source code happens to be, nobody remembers. The next hurdle is to figure out how to build and deploy it, often not documented either and nobody has done it for years. It may be tens or hundreds of thousands of lines of code, what the heck is it all supposed to do? Where might the buggy code reside in this opaque morass? An AI summary is mighty helpful in these situations even if it isn't perfectly accurate.

Yeah I have seen this whining from you previously. Clearly you don't use this tech and all you know about it is " Every AI fluffing article we've seen".

>> the closed source codebases I've seen for low quality slop

Same here, most established companies have a gigantic amount of legacy software written long ago by people who left. Little to no documentation and a rickety, brittle structure that is always teetering on failure. At least with AI you can make it document the code it writes and the architecture.

Fortunately you can also tell it to evaluate and document the legacy code. Very helpful.

I was an electronics technician back in the late 70's. I had worked for a year or so debugging circuit boards that came off of an assembly line (many parts were soldered on by hand) so I had experience with the simple IC's and CPU's of that time. This was rare skill in those days, and I leveraged that to get a job as the technician at a Computerland store that sold Apple II, Commodore Pet, Atari 400, etc. I didn't know a thing about them at first, but nobody else did either.

We didn't have circuit diagrams for most of the computers so there was little hope for repairing them, but all the IC's in the Apple were plugged in to sockets and were removable. I was able to get a diagram that showed which section of the board was responsible for what subsystem - display, keyboard, memory, I/O, etc. This made it possible to set a working machine on a bench next to a broken one and swap IC's one by one until you reached the defective component. I fixed a lot of Apples that way.

They were hugely expensive. A fully loaded Apple II cost about $2,500 in 1980, which would be about $10,000 in today's money. But people bought them! I think we sold one or two a week.

I didn't say Rocky was able to live in Earth atmosphere. I said he "swiftly sets up residence in the human starship".

In the book Grace quickly implements a language translation program on a laptop. They were communicating within hours, starting with rudimentary sign language. Meanwhile on Earth we've been largely unable to decipher the languages of fairly intelligent creatures such as birds, primates, and cetaceans. We don't understand their thought systems and vice versa, so it seems like quite a stretch to be able to almost immediately be able to communicate with a life form that isn't even water-based. I realize its essential to the story but it seems more like fantasy than science, clumsy writing.

I thought the book was badly paced, it careens from one improbable crisis to another. Even the initial setting seems artificial, all but one crew member is dead for some unexplained reason. The 'doctor' machine didn't revive them when their health started failing? And then conveniently there's an alien who is also the only survivor on his ship, strange coincidence. Grace immediately trusts this creature completely and gives it total access to his ship, highly unlikely. I can handle a few leaps of imagination but this was a pretty flimsy story.