frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

Open in hackernews

Ask HN: I hate coding agents. Is this skill issue?

3•cmar00•1h ago
1. Farewell deep focus. Before, I could work in flow state for hours on end. Now, waiting for the prompt to finish introduces constant interruptions, which translates into constant distractions, constant context switch, infinite ideas for endless new projects. My brain is melting. I can literally feel when my neurons start clogging up, forcing me to take frequent pauses just staring outside the window for 10 minutes straight.

2. I don't own my own projects anymore. Before, I used to understand every single little implementation detail of a project. When a client reported a bug, I immediately knew what and where to fix in the code base. Now, all I can do is to clasp my hands and recite a prayer to the LLM deity, hoping that my voice will be heard.

3. The overlords decide my fate. This is a corollary of #2. There's an outage every other day, and since I don't understand 80% of the code the AI has written, no AI available means it's literally impossible for me to work. I depend completely on its availability.

4. It's mostly right, but never quite there. I really make an effort to write descriptive prompts, extensive documentation, and various design docs. All of this to make it easier for the AI to understand the context and philosophy of the project. I write AGENTS.md, skills and rules. Yet, it can easily do 80% of the work correctly, but it always makes some wrong assumptions that make the other 20% very hard to fix. You have to held it's hand over every single little bug that it can't automatically fix. Old, already fixed bugs keep resurfacing out of nowhere. It's like Penelope's shroud: it looks like it fixed the current problem, but an old fix just wen't undone. No real progress was made.

This is my experience with coding agents, please tell me your own. I'd love this to be a discussion.

P.S. This post was handwritten by a human being.

Comments

ben_w•47m ago
My experience is also that they're very different from writing code myself.

There are ways to use the tool better and worse, the worst is to blindly fire off the agent and commit the changes without reading them: This *only* works for very small projects, though it often *does* work for those projects.

For bigger projects, you need to use (and enforce use of) the right architecture, and read pull requests rather than saying "LGTM" and clicking accept. Pick the right way to split the project up into discrete components with clear boundaries between them, make sure that the LLM's commits never introduce surprise coupling outside those boundaries, and make sure the unit tests aren't ridiculously brittle, they can be OK.

https://www.youtube.com/watch?v=grPtnrOTP_4

"Make sure to read the commits" seems to be enough for your points 2/3, at least for me. But everyone thinks a bit differently, YMMV. It also means I almost immediately catch most of the (20%? 10%?) of the time the LLM does something wrong, though not all of them.

I find I still have deep focus. If anything, the problem for me is they're so addictive I am tempted to keep working and working rather than rest, a problem I last faced over 20 years ago when I was working on a personal project that ended up not going anywhere.

GildenEye•30m ago
I don't really think so. When writing code with AI, I feel like we're doing two different jobs at once: writing documentation and writing code. Even though they're technically two sides of the same coin, it gets really frustrating.

Besides, no matter how good our design docs are, AI just can't grasp the business context outside of what's written. Plus, AI loves to modify things using band-aid patches, making it incredibly hard to keep the code clean. The design logic often gets messy, and you ultimately lose true logical consistency.

My current workaround is to provide detailed architecture and design principles upfront, and also add comments at the top of each file to clearly define its scope. That seems to help a bit.

But my biggest headache right now is that AI struggles to adapt to the latest APIs—it actually likes to revert the new APIs I've just updated back to the older, deprecated ones.

Vercel launching Native SDK – toolkit for building native desktop applications

https://github.com/vercel-labs/native
1•exponential_dan•2m ago•0 comments

Every JavaScript bundler handles inline <script> tags wrong

https://carter.sande.duodecima.technology/inline-script-pitfalls/
1•ptx•3m ago•0 comments

Verbalizable Representations Form a Global Workspace in Language Models

https://transformer-circuits.pub/2026/workspace/index.html
1•andsoitis•4m ago•0 comments

Calnode – self-hostable Calendly and Zoom in one Go binary

https://calnode.com/
1•shockalotti•4m ago•0 comments

SnapID – point your camera at anything, get an instant AI ID

https://apps.apple.com/us/app/snapid-object-identifier/id6761325361
1•rafaelrpadovani•5m ago•0 comments

Wall Street Banks Miss Multibillion-Dollar AI Deals in Hong Kong

https://www.bloomberg.com/news/articles/2026-07-09/wall-street-banks-miss-multibillion-dollar-ai-...
1•technewssss•5m ago•0 comments

Show HN: 18 Words

https://18words.com/
3•pompomsheep•6m ago•0 comments

A field guide to Claude Fable 5: Finding your unknowns

https://claude.com/blog/a-field-guide-to-claude-fable-finding-your-unknowns
1•maxloh•6m ago•0 comments

Generative Colors with CSS

https://gomakethings.com/articles/generative-colors-with-css/
2•surprisetalk•10m ago•0 comments

The Most Important Cheque in Economics

https://humanprogress.org/the-most-important-check-in-economics/
3•throw0101d•10m ago•0 comments

TikTok users don't have as much agency over their FYPs as they think

https://arstechnica.com/science/2026/07/how-much-control-do-tiktok-users-really-have-over-fyps/
1•derbOac•11m ago•0 comments

Pivotick is network graph library to facilitate pivoting

https://github.com/Pivotick/Pivotick
2•adulau•11m ago•0 comments

Abandoned Motorola Headquarters (2020)

https://www.abandonedspaces.com/industry/motorola.html
2•downbad_•11m ago•0 comments

Glaze: Fast, in memory, serialization, reflection, RPC library for C++

https://github.com/stephenberry/glaze
1•klaussilveira•11m ago•0 comments

Show HN: OpenDescent – a P2P messenger with no central server (libp2p, Ed25519)

https://open-descent.com
1•Jaguwa•12m ago•0 comments

Neortcw: Return to Castle Wolfenstein SP/MP/ET in a Single Build

https://github.com/klaussilveira/neortcw
1•klaussilveira•12m ago•0 comments

Databricks AI Agent Genie Code Is No Longer Free. Now You Have to Pay as You Go

https://medium.com/dev-genius/databricks-ai-agent-genie-code-is-no-longer-free-now-you-have-to-pa...
1•protmaks•13m ago•0 comments

Chocolate DOOM-3-BFG: original BFG experience, without the bugs

https://github.com/klaussilveira/chocolate-doom3-bfg
1•klaussilveira•13m ago•0 comments

I parsed 17 years of Spain's company registry – 9.5M events, no scraping

https://bormeapi.com/blog/borme-open-data-api
1•gekosp•15m ago•0 comments

Show HN: 92% of US city websites fail Ada accessibility

https://accesslumens.com/research/state-of-us-local-government-accessibility-2026
2•a11ymaster•15m ago•0 comments

What the CE mark means on your electronics

https://www.engadget.com/2209022/what-ce-marking-means-on-electronics/
3•bookofjoe•17m ago•0 comments

Ocean floor witnessed splitting apart for the first time – releasing lava

https://www.nature.com/articles/d41586-026-02139-7
2•Someone•18m ago•0 comments

Which Predictor? How GEPA Optimizes a Multi-Agent DSPy Program

https://www.elicited.blog/posts/how-gepa-optimizes-a-multi-agent-dspy-program
1•justanotheratom•19m ago•1 comments

Google DeepMind Unionization Talks Are Off to a Rocky Start

https://www.wired.com/story/google-deepmind-unionization-talks-are-off-to-a-rocky-start/
5•moxifly7•20m ago•0 comments

What goes on in a tennis player's brain facing a 238 km/h ball?

https://www.rte.ie/brainstorm/2026/0709/1582506-tennis-serve-return-brain-reaction-prediction-neu...
1•austinallegro•20m ago•0 comments

Show HN: Agent scans 130 careers pages nightly, scores jobs vs. your resume

https://github.com/tarunlnmiit/autopilot-jobhunt
1•tarunlnmiit•22m ago•1 comments

Show HN: Flow – a real-time network throughput dashboard for the terminal

https://terminaltrove.com/flow/
1•programmersd•23m ago•1 comments

A Hilariously Surprisingly Replayable Escape Room

https://mssv.net/2025/07/06/the-ladder-a-hilariously-absurd-surprisingly-replayable-escape-room/
1•surprisetalk•24m ago•0 comments

GamePassList – A fast, filterable alternative to browse Xbox Game Pass

https://gamepasslist.app/
2•jonathanfcb•26m ago•0 comments

Anthropic reveals a workspace in Claude that mirrors a theory of consciousness

https://venturebeat.com/technology/anthropics-new-j-lens-reveals-a-silent-workspace-inside-claude...
2•FrustratedMonky•27m ago•0 comments