LLMs Are Good at Some Things; Others? Yeah, Maybe Not Yet
April 16, 2026
Another dinobaby post. No AI unless it is an image. This dinobaby is not Grandma Moses, just Grandpa Arnold.
The write up “My AI Skeptic Friends Are All Nuts” is a post destined to activate the Berry-Zitron protein string. The article asserts:
Some of the smartest people I know share a bone-deep belief that AI is a fad — the next iteration of NFT mania. I’ve been reluctant to push back on them, because, well, they’re smarter than me [sic]. But their arguments are unserious, and worth confronting. Extraordinarily talented people are doing work that LLMs already do better, out of spite. All progress on LLMs could halt today, and LLMs would remain the 2nd most important thing to happen over the course of my career.
Since I am 82, I am not sure how to interpret the statement “the 2nd most important thing to happen over the course of my career.” If one is young, I suppose AI is a big deal, but I lived to watch the Pets.com-type events. I was around with IBM mainframes were the miracle machines. I also did the dad thing, and I thought that was an important thing, maybe number one.
The agent controlling other robots is not happy with the human. Glass lemons in the pitchers is an error. Meanwhile the production line hums along at high speed. Thanks, Venice.ai. Good enough.
The essay’s focus, however, is on writing code. The author is not into FORTRAN and Assembler. That’s okay, but some would argue that assembly language was a big deal, as was Unix, and the semi-conductor struck me as important too.
The write up says:
LLMs can write a large fraction of all the tedious code you’ll ever need to write. And most code on most projects is tedious. LLMs drastically reduce the number of things you’ll ever need to Google. They look things up themselves. Most importantly, they don’t get tired; they’re immune to inertia.
I agree. Certain types of code can be generated by an LLM. One example are the stealer routines useful in more complex Telegram bots. Our tests suggest that getting an LLM to handle the proprietary languages for complex TON platform activities is not ready for prime time. The idea is to split malware into chunks and have the smart contract monitor the target and deliver the chunks over time to perform the desired functions. A skilled human Telegram developer remains important. Writing a script to automate something in Word is a different type of task. Besides macros work and no smart software is needed.
The write up asserts:
Today, LLMs do a lot of typing, Googling, test cases, and edit-compile-test-debug cycles. But even the most Claude-poisoned serious developers in the world still own curation, judgement, guidance, and direction. Also: let’s stop kidding ourselves about how good our human first cuts really are.
I agree. Let smart software do the smart software part. Let a human do the human part. The only problem is that if humans don’t know what should be done and how to do it, will those humans in the loop just function like the border guard collecting money to leave Zimbabwe. Hand over cash. Open passport. Guard stamps passport and says, “Next.” Works well. Those with contraband find the system just peachy keen.
The essay points out:
If you’re taking time carefully golfing functions down into graceful, fluent, minimal functional expressions, alarm bells should ring. You’re yak-shaving. The real work has depleted your focus. You’re not building: you’re self-soothing. Which, wait for it, is something LLMs are good for. They devour schlep, and clear a path to the important stuff, where your judgement [sic and values really matter.
What about repurposing open source software? The write up says:
Meanwhile, software developers spot code fragments seemingly lifted from public repositories on Github … What about the licensing? If you’re a lawyer, I defer. But if you’re a software developer playing this card? Cut me a little slack … No profession has demonstrated more contempt for intellectual property.
I think this means that programmers as a professional group are not too concerned about intellectual property rights. If I wrong, the author presents developers of code using LLMs (many of these AI outfits are being sued for alleged IP violations) care a great deal about intellectual property. I am not sure which meaning to carry away.
The author points out:
I’m not a radical or a futurist. I’m a statist. I believe in the haphazard perseverance of complex systems, of institutions, of reversions to the mean. I write Go and Python code. I’m not a Kool-aid [sic] drinker. But something real is happening. My smartest friends are blowing it off. Maybe I persuade you. [sic] Probably I don’t. But we need to be done making space for bad arguments.
How important is AI? The answer, according to the author:
But AI is also incredibly — a word I use advisedly — important. It’s getting the same kind of attention that smart phones got in 2008, and not as much as the Internet got. That seems about right.
Net net: AI is a tool, a utility. Ask it to do something else and surprises may manifest. My concern is that CEOs with MBAs and bean counters may see a simpler line connecting LLM to cost reduction. That’s the sporty part.
Stephen E Arnold, April 16, 2026
Comments
Got something to say?
-
- Subscribe to Beyond Search
Feature archive
News archive
-
