frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

Open in hackernews

The current state of LLM-driven development

http://blog.tolki.dev/posts/2025/08-07-llms/
40•Signez•4h ago

Comments

randfish•2h ago
Deeply curious to know if this is an outlier opinion, a mainstream but pessimistic one, or the general consensus. My LinkedIn feed and personal network certainly suggests that it's an outlier, but I wonder if the people around me are overly optimistic or out of synch with what the HN community is experiencing more broadly.
Terretta•2h ago
Which part of the opinion?

I tend to strongly agree with the "unpopular opinion" about the IDEs mentioned versus CLI (specifically, aider.chat and Claude Code).

Assuming (this is key) you have mastery of the language and framework you're using, working with the CLI tool in 25 year old XP practices is an incredible accelerant.

Caveats:

- You absolutely must bring taste and critical thinking, as the LLM has neither.

- You absolutely must bring systems thinking, as it cannot keep deep weirdness "in mind". By this I mean the second and third order things that "gotcha" about how things ought to work but don't.

- Finally, you should package up everything new about your language or frameworks since a few months or year before the knowledge cutoff date, and include a condensed synthesis in your context (e.g., Swift 6 and 6.1 versus the 5.10 and 2024's WWDC announcements that are all GPT-5 knows).

For this last one I find it useful to (a) use OpenAI's "Deep Research" to first whitepaper the gaps, then another pass to turn that into a Markdown context prompt, and finally bring that over to your LLM tooling to include as needed when doing a spec or in architect mode. Similarly, (b) use repomap tools on dependencies if creating new code that leverages those dependencies, and have that in context for that work.

I'm confused why these two obvious steps aren't built into leading agentic tools, but maybe handling the LLM as a naive and outdated "Rain Man" type doesn't figure into mental models at most KoolAid-drinking "AI" startups, or maybe vibecoders don't care, so it's just not a priority.

Either way, context based development beats Leroy Jenkins.

WD-42•1h ago
I think it’s pretty common among people whose job it is to provide working, production software.

If you go by MBA types on LinkedIn that aren’t really developers or haven’t been in a long time, now they can vibe out some react components or a python script so it’s a revolution.

danielbln•1h ago
Hi, my job is building working production software (these days heavily LLM assisted). The author of the article doesn't know what they're talking about.
MobiusHorizons•1h ago
My impression has been that in corporate settings (and I would include LinkedIn in that) AI optimism is basically used as virtue signaling, making it very hard to distinguish people who are actually excited about the tech from people wanting to be accepted.

My personal experience has been that AI has trouble keeping the scope of the change small and targeted. I have only been using Gemini 2.5 pro though, as we don’t have access to other models at my work. My friend tells me he uses Claud for coding and Gemini for documentation.

ebiester•1h ago
I disagree from almost the first sentence:

> Learning how to use LLMs in a coding workflow is trivial. There is no learning curve. You can safely ignore them if they don’t fit your workflows at the moment.

Learning how to use LLMs in a coding workflow is trivial to start, but you find you get a bad taste early if you don't learn how to adapt both your workflow and its workflow. It is easy to get a trivially good result and then be disappointed in the followup. It is easy to try to start on something it's not good at and think it's worthless.

The pure dismissal of cursor, for example, means that the author didn't learn how to work with it. Now, it's certainly limited and some people just prefer Claude code. I'm not saying that's unfair. However, it requires a process adaptation.

mkozlows•1h ago
"There's no learning curve" just means this guy didn't get very far up, which is definitely backed up by thinking that Copilot and other tools are all basically the same.
leptons•45m ago
Basically, they are the same, they are all LLMs. They all have similar limitations. They all produce "hallucinations". They can also sometimes be useful. And they are all way overhyped.
tptacek•1h ago
Learning how to use LLMs in a coding workflow is trivial. There is no learning curve. You can safely ignore them if they don’t fit your workflows at the moment.

I have never heard anybody successfully using LLMs say this before. Most of what I've learned from talking to people about their workflows is counterintuitive and subtle.

It's a really weird way to open up an article concluding that LLMs make one a worse programmer: "I definitely know how to use this tool optimally, and I conclude the tool sucks". Ok then. Also: the piano is a terrible, awful instrument; what a racket it makes.

edfletcher_t137•1h ago
The first two points directly contradict each other, too. Learning a tool should have the outcome that one is productive with it. If getting to "productive" is non-trivial, then learning the tool is non-trivial.
prerok•54m ago
I agree with your assessment about this statement. I actually had to reread it a few times to actually understand it.

He is actually recommending Copilot for price/performance reasons and his closing statement is "Don’t fall for the hype, but also, they are genuinely powerful tools sometimes."

So, it just seems like he never really gave a try at how to engineer better prompts that these more advanced models can use.

dezmou•1h ago
OP did miss the vscode extension for claude code, it is still terminal based but: - it show you the diff of the incoming changes in vscode ( like git ) - it know the line you selected in the editor for context
sudhirb•1h ago
I have a biased opinion since I work for a background agent startup currently - but there are more (and better!) out there than Jules and Copilot that might address some of the author's issues.
simonw•52m ago
Learning how to use LLMs in a coding workflow is trivial. There is no learning curve. [...]

LLMs will always suck at writing code that has not be written millions of times before. As soon as you venture slightly offroad, they falter.

That right there is your learning curve! Getting LLMs to write code that's not heavily represented in their training data takes experience and skill and isn't obvious to learn.

Show HN: The current sky at your approximate location, as a CSS gradient

https://sky.dlazaro.ca
373•dlazaro•6h ago•79 comments

Long-term exposure to outdoor air pollution linked to increased risk of dementia

https://www.cam.ac.uk/research/news/long-term-exposure-to-outdoor-air-pollution-linked-to-increased-risk-of-dementia
184•hhs•7h ago•55 comments

MCP's Disregard for 40 Years of RPC Best Practices

https://julsimon.medium.com/why-mcps-disregard-for-40-years-of-rpc-best-practices-will-burn-enterprises-8ef85ce5bc9b
154•yodon•5h ago•76 comments

Simon Willison's Lethal Trifecta Talk at the Bay Area AI Security Meetup

https://simonwillison.net/2025/Aug/9/bay-area-ai/
157•vismit2000•5h ago•53 comments

Debian 13 "Trixie"

https://www.debian.org/News/2025/20250809
197•ducktective•1h ago•67 comments

A CT scanner reveals surprises inside the 386 processor's ceramic package

https://www.righto.com/2025/08/intel-386-package-ct-scan.html
74•robin_reala•3h ago•13 comments

OpenFreeMap survived 100k requests per second

https://blog.hyperknot.com/p/openfreemap-survived-100000-requests
265•hyperknot•6h ago•63 comments

Quickshell – building blocks for your desktop

https://quickshell.org/
181•abhinavk•4d ago•28 comments

ESP32 Bus Pirate 0.5 – A Hardware Hacking Tool That Speaks Every Protocol

https://github.com/geo-tp/ESP32-Bus-Pirate
65•geo-tp•5h ago•12 comments

Stanford to continue legacy admissions and withdraw from Cal Grants

https://www.forbes.com/sites/michaeltnietzel/2025/08/08/stanford-to-continue-legacy-admissions-and-withdraw-from-cal-grants/
97•hhs•7h ago•159 comments

The current state of LLM-driven development

http://blog.tolki.dev/posts/2025/08-07-llms/
40•Signez•4h ago•15 comments

ChatGPT Agent – EU Launch

https://help.openai.com/en/articles/11752874-chatgpt-agent
52•Topfi•4h ago•35 comments

Accessibility and the Agentic Web

https://tetralogical.com/blog/2025/08/08/accessibility-and-the-agentic-web/
14•edent•3h ago•5 comments

Testing Bitchat at the music festival

https://primal.net/saunter/testing-bitchat-at-the-music-festival
37•alexcos•3d ago•22 comments

The mystery of Alice in Wonderland syndrome

https://www.bbc.com/future/article/20230313-the-mystery-of-alice-in-wonderland-syndrome
11•amichail•3d ago•5 comments

Isle FPGA Computer: creating a simple, open, modern computer

https://projectf.io/isle/fpga-computer.html
9•pabs3•3d ago•0 comments

Jan – Ollama alternative with local UI

https://github.com/menloresearch/jan
148•maxloh•10h ago•65 comments

Ratfactor's Illustrated Guide to Folding Fitted Sheets

https://ratfactor.com/cards/fitted-sheets
93•zdw•8h ago•13 comments

Knuth on ChatGPT (2023)

https://cs.stanford.edu/~knuth/chatGPT20.txt
90•b-man•3h ago•29 comments

End-User Programmable AI

https://queue.acm.org/detail.cfm?id=3746223
25•tosh•4h ago•1 comments

Cordoomceps – replacing an Amiga's brain with Doom

https://mjg59.dreamwidth.org/73001.html
25•naves•4d ago•4 comments

Ch.at – a lightweight LLM chat service accessible through HTTP, SSH, DNS and API

https://ch.at/
5•ownlife•1h ago•0 comments

I want everything local – Building my offline AI workspace

https://instavm.io/blog/building-my-offline-ai-workspace
976•mkagenius•1d ago•261 comments

The dead need right to delete their data so they can't be AI-ified, lawyer says

https://www.theregister.com/2025/08/09/dead_need_ai_data_delete_right/
137•rntn•6h ago•94 comments

Car has more than 1.2M km on it – and it's still going strong

https://www.cbc.ca/news/canada/nova-scotia/1985-toyota-tercel-high-mileage-1.7597168
164•Sgt_Apone•3d ago•213 comments

Sandstorm- self-hostable web productivity suite

https://sandstorm.org/
142•nalinidash•14h ago•28 comments

Mexico to US livestock trade halted due to screwworm spread

https://www.usda.gov/about-usda/news/press-releases/2025/07/09/secretary-rollins-takes-decisive-action-and-shuts-down-us-southern-border-ports-livestock-trade-due
214•burnt-resistor•5h ago•169 comments

Physical Media Is Cool Again. Streaming Services Have Themselves to Blame

https://www.rollingstone.com/culture/culture-features/physical-media-collectors-trend-viral-streamers-1235387314/
18•coloneltcb•1h ago•7 comments

How AI is upending the software development industry

https://www.reuters.com/lifestyle/bootcamp-bust-how-ai-is-upending-software-development-industry-2025-08-09/
5•wglb•2h ago•0 comments

Residents cheer as Tucson rejects data center campus

https://www.datacenterdynamics.com/en/news/residents-cheer-as-tucson-rejects-amazons-massive-project-blue-data-center-campus-in-arizona/
36•01-_-•2h ago•18 comments