frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

Ask HN: Have LLMs Plateaued?

4•leandrobon•2h ago•5 comments

Tips for Landing a YC Internship?

6•cnnadozi•5h ago•3 comments

Ask HN: What are you working on? (August 2026)

12•sarreph•6h ago•19 comments

Ask HN: What's Next for LLMs?

4•apatheticonion•4h ago•2 comments

What Is Decispher?

3•cool_coder12•7h ago•1 comments

Ask HN: Anyone still do work on Intel Macs?

20•senor_digimon•1d ago•27 comments

Best way to avoid bloat and AI – selfcontaining OS

11•bialamusic•16h ago•2 comments

Tell HN: Amazonbot aggressively scraping my website and ignoring robots.txt

19•pera•1d ago•17 comments

Ask HN: I still don't understand why AI agents need "skills"

14•skeptic_ai•1d ago•13 comments

Ask HN: What's Your BHAG?

7•sudo_cowsay•16h ago•7 comments

What is the downside of coaching that I may not be seeing yet?

4•ikut3hva•17h ago•0 comments

Ask HN: How are you using AI to learn?

8•mfalcon•1d ago•17 comments

Ask HN: When will ChatGPT Voice be able to control my phone?

2•spapas82•19h ago•6 comments

What if we learned languages like we acquired our first?

7•Salim_wariz•20h ago•9 comments

ask HN: How do I run Android 15 on a system with Andr 12 vendors?

3•KinetiNode•21h ago•2 comments

Tell HN: HackerOne Announces Mandatory identity verification

17•0xmattf•2d ago•1 comments

Strategic Oil Reserves Down to a Record Low Two-Week Supply

20•megamike•3d ago•20 comments

Ask HN: What did you learn last month? (July 2026)

15•bhu1st•2d ago•14 comments

AI will always find a difference

3•ghassenfaidi•1d ago•1 comments

Ask HN: Best Blogging Platform?

5•asxndu•1d ago•6 comments

Ask HN: How did your boss's behavior change with the rise of AI?

8•aredirect•1d ago•9 comments

Which web browser has no AI?

10•lilerjee•1d ago•13 comments

Ask HN: Should organizations be allowed to have all types of personl data(X.com)

4•dfps•1d ago•0 comments

Ask HN: When did we go from agentic loops to graphs?

4•grandimam•1d ago•2 comments

You've reached the end!

Open in hackernews

Ask HN: What's Next for LLMs?

4•apatheticonion•4h ago
I've been using DeepSeek v4 flash for guided agent workflows (in-IDE, prompt/review diffs) and have found no noticeable benefit to using Haiku, Opus, Sonnet in this workflow.

For around a dollar a day, I am able to more than double my own productivity. What are the frontier models for/what are they trying to solve?

Are they designed to one shot applications? Is the expectation that we want to be able to "yolo" prompt LLMs and have them complete work unsupervised?

Comments

FergusArgyll•2h ago
If you can't find a difference between Haiku & Opus, something's going wrong
apatheticonion•2h ago
It's because of my workflow, I think.

I have my IDE open and a prompt a given task with limited scope - like "write a function that does xyz" "add tests" "create an endpoint that accepts this json" "write a sql repository for this schema" etc.

I use Rust on the backend so if it compiles it works, meaning I don't need to worry about runtime errors - if it looks right, it's right. On the frontend I use Angular, which is about as constrained an environment as you can get in the FE world, so it takes a lot to lead an LLM astray.

So my requirement for high intelligence is pretty low and I am able to review diffs quickly. Deepseek flash is extremely fast so it can render my prompts quickly - while Haiku, Opus and Sonnet are slow and create the same outputs.

I'm guessing the value of Sonnet+ is less hand holding letting you give it larger scopes of work? Giving an AI a high level task and letting it do its thing without needing to review the output?