I always find so much cognitive dissonance in these kinds of posts.

> My most successful work this year was hand-writing a tool, replacing ~4,000 lines of vibe-code with a few hundred. My system has proven more flexible, easier to test, and more generalizable than what it replaced.

> agents still need human taste and experience to produce good work.

Where do you think human taste and experience in software comes from?

> overstimulating & exhausting in a way that coding never was.

> for many of us, honing the code, finding the right design, seeing the parts mesh together — that was the meditative purpose we found in our jobs.

> That’s not the only time AI waylaid me. LLMs could not figure out how to annotate that model’s training data correctly. I spent days going back and forth with them

> LLMs are not yet good companions when brainstorming complicated things.

Why do we expect this to get better? The training data will degrade in quality if more people rely on LLMs so I'd expect models to eventually start getting worse even if they currently enjoy apparent improvements in output.

> Writing code is solved.

What a strange thing to say given all the above experiences, which I think we all share.

Why are we willing to give up so much in terms of purpose and understanding for perceived productivity gains?

I believe that anyone who says that writing code is no longer needed, was probably not a very good coder to begin with. I have seen what the best models can do, and I've seen what people call "good", and I am not impressed.

However, I'm still split. Because this could mean that we see a jump in code quality, due to bad coders getting someone better to code for them. But we could see a decrease in code quality, due to less people "caring".

There is no substitute for reading the code in many many cases. I spend more than 3k/mo on tokens. I find incorrect assumptions baked into the code (the ultimate spec) and missed by the agents and tests daily. This is more pronounced in data analysis and complex systems.

As a side effect having to read the code at least once inevitably makes it way shorter and usually much better.

> I'm happy to test crazy new ideas and throw out what doesn’t work well. It was able to wire & test a huge change to Mumble that added better noise cancelling; I use my beta build of it nearly every day.

To me, this is a negative and mostly irresponsible use of AI.

OP dumped a huge PR (edit: not actually huge) on the Mumble maintainers,[0] which didn't match project conventions. He left a bunch of unresolved conversations with Claude that are hard to follow because it looks like an insane person having a conversation with themselves.

Then, the Mumble maintainers gave him pretty polite, patient feedback, and he just ignored it. No follow up for going on three months. Granted, the maintainers left the PR in the queue for four months, so it's not like OP is the slowest link, but I definitely wouldn't point to this as a win when there's basically no traction on the PR or any indication this will be merged.

[0] https://github.com/mumble-voip/mumble/pull/7180

Edit: radlad is correct that the PR is not "huge" at ~300 LOC. I was reading on my phone and didn't check the PR size and mistakenly trusted OP's assessment of "huge."

OP: [last year][1], I shared how I was using AI day-to-day as an ordinary engineer. This year, I did it again!

It's crazy realizing that Claude Code only started getting big in the last year or so. It feels so different now, with all the new models & tools we have to use. I hope you enjoy!

[1]: https://ben.stolovitz.com/posts/how_use_ai_oct_2025

>I spend about ~$3,000/ month in tokens.

For whom thinks that there is no economic value in local AI stations beside privacy, availability, transparency, unresrtictibility etc...

Whoever says they throw agents at work they (the human) are getting paid to do, i am very curious about how you make sure the thing that is being built is it's money worth.

Every time I see a blog about how coding is "obsolete", it is written by a person literally working in AI. They are incentivized to push this narrative due to their job...

Why is product search still so bad? Google should have solved it ages ago, but even with AI it is still bad.

Nothing against the post itself, but something I find funny: the development in AI is so rapid, that everyone thinks they are ahead of everybody else. Here we have a blog post telling us "coding is solved". In Oct 2026, how many of us don't know this already?

3,000 USD a month on tokens seems huge to me! I am on a Claude Code Team standard plan (15 GBP a month) and it is rare that I hit my usage limits even though I am using it five days a week for coding

AI, like people, seems to get better as experience builds up. (thread context) But that experience is getting used to the workflow that fits that user, so it seems hard to know in advance. So, like the post said, since you only learn the better approach or structure "after doing it", this seems like a pretty common situation.

> I also now reckon with the fatigue of using these agents. Managing 6 different conversations with 6 different robots about 6 different features is overstimulating & exhausting in a way that coding never was.

Big +1 to that.

its like the seventh time they solved code, do you think its a good solution this time? i want to retire and paint waifus already

> The biggest change has been Google’s AI summary. Google’s first or second result is almost always an AI summary, and it’s usually decent.

Honestly this makes me skeptical of the whole piece. I get a pretty strong Gell-Mann amnesia effect from it[0]. It is not quite reliable enough that you can trust it, as when it is wrong it tends to be VERY wrong. In addition it is not like 99% accurate or something, but closer to like 80-90% or something.

[0]: The phenomenon of a person trusting newspapers for topics which that person is not knowledgeable about, despite recognizing the newspaper as being extremely inaccurate on certain topics which that person is knowledgeable about.

The "Art & music & video" section is really puzzling to me. On one hand, the author mentions liking some AI music, and then links to low-effort songs that at best you could call sort-of catchy... and on the other hand, the author writes about AI video: "Has it ever been used for good?"

Gen-AI can and is being used to make real, worthwhile movies. The best video model available, Seedance 2.5, can't produce clips longer than 30 seconds. The human artistry comes from which clips/shots you ask for and how you edit them together to tell a story. Plenty of people are doing good work here.

Here are some examples. Even if you don't like these personally, they aren't slop. You can feel the creator's intention behind them:

https://www.youtube.com/watch?v=2__VhU7_3lw

https://www.youtube.com/watch?v=gl3oob03rdE

https://higgsfield.ai/@nue_nue/projects/listen-to-meh

https://higgsfield.ai/@sssacccade/projects/hunger-ai-short-f...

(This one is mine, just FYI) https://www.youtube.com/watch?v=g8XUH2OXXfA

> Open-source models already outperform those. If OpenAI and Anthropic start pricing me out of their top-tier models, I’ll be able to buy a few GPUs and hack on personal projects with some local agent buddies. So that’s at least one doom scenario I’m not super worried about.

Does anyone find this to be true?

The models can literally get any Jira ticket and one-shot, but we still have demand for SWE, so I don't get people saying it's gonna "take our jobs".

LLMs are just a tool, you can one-shot a full project, but you still need an engineer to validate, it will catch some adjustments, then more tickets will be created from that and the cycle goes on and on.

$3,000/month is nuts! For $36k (1 yr of spend) you can buy a system with dual rtx pro 5000 (144GB VRAM). or spend a little more ($44k) and get two M5 Ultras w/512GB RAM. the latter, clustered, is big enough to run the biggest open weights (quantized), or smaller MoE weights on each in parallel. what you get for these systems is way more than $3k/mo of tokens, because you can run it 24/7/365. you can have pass after pass after pass of optimization, bug fixing, orchestrated subagents on backlogs, etc. way more value than paying for managed inference

> I also now reckon with the fatigue of using these agents. Managing 6 different conversations with 6 different robots about 6 different features is overstimulating & exhausting in a way that coding never was.

Anyone using a swarm of agents in November 2026 to write code is no longer at the frontier of capability.

Running multiple system prompts at once is a perfect example of anthropomorphism. We have this fantasy that a community of diversely aligned agents will somehow outperform a single well-aligned agent. There might be some cases where the random walk takes you where you want to go, but how stable and stressful do you want it to be?

Some of us are using AI as a means to make a living now. If I slip even a little bit on best practices my clients will likely turn into ghosts. Multi agent is an extremely expensive form of entertainment in this context. It's the entire basis for why we have these junkies trolling around HN complaining about how they can never get enough tokens. If I was still playing with these ideas in front of my customers I can guarantee I'd be 100% unemployed right now. They brought me in to make AI into a profit center, not a cost center.

> Video generation Has it ever been used for good?

I guess he hasn't seen CRAFT (1979) yet, which would not have been possible before AI.

Not that it redeems all the other AI thingies, it's always a mixed bag, yet I can't say a single bad thing about it.

I find it amusing, that

> my uncle can create software

Is considered great

> my uncle can create music or art

Is not considered great.

I guess humans have agreed, that art is a skill you must learn if you want to produce it, but engineering or software is not a skill worth learning and we are happy if computers can do that for us.

So it’s great and amazing if anyone can create their gizmos and software to help them but not if they create music or digital images, which apparently it is wrong to prompt, and not a sign of a great future.

> I still believe I use AI fairly typically

...

> Last October, I was an engineer on Microsoft’s Windows AI team. I am now a research engineer at Salesforce AI Research.

That's generally not how "typical use" works.

TIL: AI Doompop

Always good to see how others use the same tools you do

> since agents write everything, it is simple to propose sweeping changes to my teammates during code review, or even write my own versions — both things unthinkable in the era of hand-written code.

That is both untrue, and should be something that occurs rarely.

Past Mid level, most dicussion in code review is centered around the conceptual level. Code is mostly use to highlight the point being made. I’ve never seen the need for someone to rewrite a patch proposed by another person. Add to it or correct misunderstandings, but never sweeping changes.

>I spend about ~$3,000/ month in tokens.

...

>In case you’re wondering, I’m subscribed to: > Claude Pro ($17/ mo) > GitHub Copilot Pro ($10/ mo)

Nothing wrong there at all, amirite?

Maybe AI scrapped basic arithmetic and now those numbers are totally fine

>I’m really enjoying some AI music

I think I would rather shoot myself in the head

Interesting article! I especially liked the AI music part.

How I use vim and look at my `.vimrc` of 2026.

OK, yes, I am not a pro dev... still, that guy blows more tokens a month then I have consumed since the conception of GPT3! Granted, I stopped using agents a few months ago, when I got my first "you reached 80% of your monthly limit" message. Since then, I am manually constructing the context and executing edits... which brought my token bill down at least 5x and gave me 2x more useable code. But yeah, I am sure, 3k$ tokens a month is not wasteful behaviour at all :-)

I find that AI should be used more extensively for System Design. I have seen a lot of people not use AI for that and come up with frighteningly complex designs and let their agents run loose to implement it.

Retroactively speaking, I engaged with the already agreed designs and used AI to find much better ways to do the same thing.

Overall, I wish people used AI even for System Design as that seems to have a higher impact.

Nice post!

It’s kinda nice seeing more nuanced takes from folks as the bubble kinda deflates and both sides realize their idealized vision isn’t coming to pass. AI isn’t going to replace all the jobs, but it’s also not going away; finally, it is merely a tool in our toolchest.

What does amuse me to no end, however, is how many craftsfolk in AI circles are loudly rejecting AI imagery and video in favor of humans. The only folks who seem fine with that are the grifters looking to slop out a bunch of adverts quickly for commercial gain rather than do actual art, or folks who want “art” but don’t want to pay a human to make it for them or learn the skills themselves. Both of those camps are viewed with disdain and loathing by both sides of the AI Boom/Doom arguments, so I’ll take whatever allies I can get on those fronts.

AI music…man, I live with artisans and musicians and we have a number of “artists” we suspect are AI but whose music is pretty solid. We’re waiting for the shoe to drop, but I’m trying to look at it as a kind of turbocharged Fruity Loops/Logic as opposed to a human replacement. I hope I’m right on that, but time will tell.

The game changer I think more doomers need to experience is agents. It sounds like a recipe for disaster (a bunch of models known to hallucinate and allowed to run roughshod over a computer unattended? What could go wrong?), and it often (hilariously) is for the scammers out there giving it full access to production and fucking off to hawk their latest Masterclass on X or LinkedIn, but that first time you give it a reasonably achievable prompt and come back to a halfway decent product you can review and refine into production in 15m or less?

Absolutely magical, doubly so when it’s just 27B Qwen via Hermes and not some souped up stack in a frontier lab costing you thousands a month in tokens. If I were to evangelize AI, that’s how I’d do it. Fuck subscriptions, fuck cloud compute, fuck frontier labs and token buckets and whatever slop Sundar/Scam/Dario/Elmo et al are hawking this week; run that locally, give it a reasonable prompt that solves a real problem, and show them the results.

Then do the right thing and go for a walk with the two hours of free time the AI just freed up for you. Don’t ask for permission to take back your time.

On the one hand, I like this post: Like developing a personal Mission Statement (one of the best pieces of advice a mentor gave me long, long ago, in a galaxy far, far away). But on the other, this sounds like the beginning of AI psychosis, or somebody that got really drunk one night and decided to blog about how amazing compilers are: "I love you, man!"

That being said, I agree with this statement: "I believe they are the greatest teachers humanity has ever created." LLMs are exceedingly powerful at accelerating the learning process.

> But I’m more and more excited about using LLMs as teachers to explore well-known concepts.

Me too! I've done several deep dives in various topics just in the last few weeks. I start off with some question I want answers to, and often end up with more questions along the way and I end up learning a ton.

For example I've been enthusiastic about nuclear energy for a long time and I've been wondering why don't we build more of it? It seems like such a great investment, the fuel is extremely cheap compared to the power it produces, even including disposal of waste and everything it seems to me like the best power source we have by far.

I've read things like "it's not economically viable" which I didn't really get - how can it both be cheap clean electricity and not economically viable? It turns out the problem is interest. If you spend $20B building a nuclear plant, you need to make enough money to pay the interest on that $20B. The plant itself will easily make much more than $20B over its lifetime but it won't cover the loan payments unless you can secure a contract with a customer willing to pay a very premium price for stable power for like 20 years.

On the other hand, China builds nuclear plants for like $3B which changes everything. They can do that because they've built lots of nuclear plants, so they have repeatable designs and established supply chains, experienced builders etc. We could do that too and I really think we should (we as in the rest of the world), I think we should invest in nuclear technology to build experience and supply chains etc in order to be able to make the technology economically viable for us. It seems to me that fusion will have exactly the same problem, if we ever even make it work. Good luck building "economically viable" fusion plants, they will certainly be a thousand times more complex than fission plants.

Anyway this wasn't supposed to be a comment about nuclear energy. I also did a deep dive on hydropower based on some claims I heard and learned lots of cool things about that. The LLM found lots of interesting source material for me, including a letter from our ministers of energy and finance talking about a proposed law change that might take effect next year which I would never normally read but since it basically answered many of the questions I had I ended up reading the whole thing and it was super interesting. It's also obviously a lot more credible than just the AI's claims.

Highly recommend trying it out about any topic you may be interested in. I use the free version of ChatGPT. I've also used the Google search AI a lot, I'll basically just Google a question and then proceed to ask followup questions etc.

I'm sure it's not all 100% correct but I'm also quite sure that most of it is pretty good quality, and it really helps when it includes links to sources, I often read the sources instead of the AI summary.

you can totally tell Claude to take some AI technical writing (which is, to be clear, writing where you explicitly need to not sound creative or interesting, and you need to maintain very strict "voice" [a term I only even know because claude explained it to me] so this is not a creative endeavor) and make it sound "less AI". At least working with opus 5.5, textual output already has a lot fewer of the colons and "load bearing" of a few months ago, and if a PR does have that kind of lingo I can just have Claude rewrite it "like a human" and it does great.

[flagged]

[flagged]

[flagged]

HN title: I use AI

> I still believe I use AI fairly typically, but I’ve had a notable change in background. Last October, I was an engineer on Microsoft’s Windows AI team. I am now a research engineer at Salesforce AI Research.

Priest wears robes.

I guess it’s one of those nonsensical automatic rules.

[dead]

[flagged]

[dead]

[flagged]

[dead]

[dead]

[flagged]

[flagged]