Hacker News Reader

Comment by DaedalusII | Hacker News Reader

If you include microsoft copilot trials in fortune 500s, absolutely. A lot of major listed companies are still oblivious to the functionality of AI, their senior management don't even use it out of laziness

View on HN

There's a lot of rote work in software development that's well-suited to LLM automation, but I think a lot of us overestimate the actual usefulness of a chatbot to the average white-collar worker. What's the point of making Copilot compose an email when your prompt would be longer than the email itself? You can tell ChatGPT to make you a slide deck, but slide decks are already super simple to make. You can use an LLM as a search engine, but we already have search engines. People sometimes talk about using a chatbot to brainstorm, but that seems redundant when you could simply think, free from the burden of explaining yourself to a chatbot.

LLMs are impressive and flexible tools, but people expect them to be transformative, and they're only transformative in narrow ways. The places they shine are quite low-level: transcription, translation, image recognition, search, solving clearly specified problems using well-known APIs, etc. There's value in these, but I'm not seeing the sort of universal accelerant that some people are anticipating.

That's probably true for some, but I think a lot of big orgs are simply risk-averse and see AI in general as a giant risk that isn't even fully baked enough to quantify yet. The security and confidentiality issues alone will make Operations hesitant, and Legal probably has some questions about IP (both the risk of a model outputting patented or otherwise protected code, and the huge legal gray area that is the copyrightability of the output of an LLM).

Give it a year or two and let things settle down and (assuming the music is still playing at that time) you might see more dinosaurs start to wander this way.

it turns out it's really hard to get a man to fish with a pole when you don't teach them how to use the reel

Or give them a stick with twine and a plastic fork as a hook, as is the case with Copilot.

If AGI is coming, won't there just be autofishers and no one will ever have to fish again, completely devaluing one's fishing knowledge and the effort put in to learn it?

It’s not a great analogy but...

“Autofishers” are large boats with nets that bring in fish in vast quantities that you then buy at a wholesale market, a supermarket a bit later, or they flash freeze and sell it to you over the next 6-9 months.

Yet there’s still a thriving industry selling fishing gear. Because people like to fish. And because you can rarely buy fish as fresh as what you catch yourself.

Again, it’s not a great analogy, but I dunno. I doubt AGI, if it does come, will end up working the way people think it will.

> If AGI is coming

spoiler, it's not

Evey now model is better than a previous one. When and why it is going to stop? Current models are limited by DC capacity which is rapidly expanded. Genuinely curious, I don't know the answer.

In regards to copilot, they’ve also been led on a fishing expedition to the middle of a desert

I'm not sure if this was the intention of the analogy, but fishing poles don't have reels.

100% All of the people who are floored by AI capabilities right now are software engineers, and everyone who's extremely skeptical basically has any other office job. On investigating their primary AI interaction surface, it's Microsoft Co-Pilot, which has to be the absolute shittiest implementation of any AI system so far. As a progress-driven person, it's just super disappointing to see how few people are benefiting from the productive gains of these systems.

This isn't my experience. I see many non-software people using AI regularly. What you may be seeing is more: organizations with no incentive to do things better never did anything to do things better. AI is no different. They were never doing things better with pencil and paper.

I think anthropic will succeed immensely here because when integrated with Microsoft365 and especially Excel it basically does what co-pilot said it would do.

The moment of realisation happen for a lot of normoid business people when they see claude make a DCF spreadsheet or search emails

claude is also smart because it visually shows the user as it resizes the columns, changes colours, etc. Seeing the computer do things makes the normoid SEE the AI despite it being much slower

No one wants a chatbot “integrated” with excel and office365 crap, it’s clippy 2.0 bullshit.

Replace excel and office stuff with ai model entirely then people will pay attention.

that only works if you can oneshot. but nobody can oneshot.

iterating over work in excel and seeing it update correctly is exactly what people want. If they get it working in MSWord it will pick up even faster.

If the average office worker can get the benefit of AI by installing an add-on into the same office software they have been using since 2000 (the entire professional career of anyone under the age of 45), then they will do so. its also really easy to sell to companies because they dont have to redesign their teams or software stack, or even train people that much. the board can easily agree to budget $20 a head for claude pro

the other thing normies like is they can put in huge legacy spreadsheets and find all the errors

Microsoft365 has 400 million paid seats

> normoid

Do you work extra hard to be this arrogant or does it come naturally?

Hilarious lack of self awareness. Calling others "normoids" yet you believe you can emphasise with them enough to predict how they will adopt AI?

I'm a SWE who's been using coding agents daily for the last 6 months and I'm still skeptical.

For my team at least, the productivity boost is difficult to quantify objectively. Our products and services have still tons of issues that AI isn't going to solve magically.

It's pretty clear that AI is allowing to move faster for some tasks, but it's also detrimental for other things. We're going to learn how to use these tools more efficiently, but right now, I'm not convinced about the productivity gain.

> I'm a SWE who's been using coding agents daily for the last 6 months and I'm still skeptical.

What improvements have you noticed over that time?

It seems like the models coming out in the last several weeks are dramatically superior to those mid-last year. Does that match your experience?

Not the grandparent, but I've used most of the OpenAI models that have been released in the last year. Out of all of them, o3 was the best at the programming tasks I do. I liked it a lot more than I like GPT 5.2 Thinking/Pro. Overall, I'm not at all convinced that models are making forward progress in general.

Yes, it matches my experience. Now I can throw tasks at the agent and have it write a full PR, with tests, good summary. Or it can review things and make good suggestions that a casual reviewer or non-expert would have missed. It can also take a bunch of logs as input, find the issue, fix the code. I can't deny it's impressive and useful.

What I'm still skeptical about is how much more productive it makes us. In my case, coding is maybe 50% of my job, and I work on complex and novel systems. The agent gives me the illusion I don't need to think anymore, but it's not the case. Agents slow me down in many cases too, I'm not learning and improving as I used to.

In a team of one at work I see clear benefits, but having worked in many different team sizes for most of my career I can see how it quickly would go down, especially if you care about quality. And even with the latest models it’s a constant battle against legacy training data, which has gotten worse over time. ”I have to spend 45 minutes explaining why a one minute AI generated PR is bad code” was how an old colleague summarized it.

Is your backlog and/or your velocity increasing, decreasing, or the same? That's really the ultimate question.

IMO Copilot was "we need to give these people rope, but not enough for them to hang themselves". A non technical person with no patience and access to a real AI agent inside a business is a bull in a china shop. Copilot Cowork is the closest thing we have to what Copilot should have been and is only possible now because models finally got good enough to be less supervised.

FWIW Gemini inside Google apps is just as bad.

Reply on news.ycombinator.com