frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

fp.

Open in hackernews

Ask HN: Ten YoE as a software engineer, what's next with LLMs coming in fast?

4•Tmkly•1h ago
I am a software engineer with 10 years of experience, currently working as a Lead React Native Engineer for a fin-tech app in London. I think it's pretty clear that LLMs are gonna change the industry massively. While I'm not immediately worried about my future career, I think in the next 5 years there will be a huge squeeze as LLMs take up more and more of the work load. I love writing code and solving problems, and I don't want to just be an LLM babysitter, and I'd like to think I'm pretty good at my job. And I want to stay technical too, so no woodworking suggestions.

So over the past month or two I've been contemplating what's next for me. I really enjoy working with React Native, Typescript etc but LLMs are pretty good at these things already (largely due to the amount of open-source JS online already). For example, If I was to transition in to some kind of AI engineering/research, where would I start with that? Or also I was thinking about embedded software engineering - again, where do I start with that?

Or am I just being too pessimistic? I do think the whole 'AI' thing is a bit overblown, so maybe it'll all be fine :shrug:.

Comments

osivertsson•1h ago
Understand the business thoroughly to know what value you create. Transition within that framework, where in this organization and with what tools can I provide more value?

If you take a moral stand against LLMs understand that your relative value will decrease.

However, with 10 YoE you should be able to quickly evaluate code and ideas proposed by LLMs for correctness and fit into your code-base. In this way you are in a better spot than junior devs.

Limit complexity. LLMs sure allows complexity to get out of control, especially when people stop caring and just wants to get things out of their hands quickly.

If you really were to transition somewhere else it has to be to something you really feel the urge to do because it gives you a real sense of joy or fascination. Something you can’t resist dabbling with already today. From your message I can tell it is not embedded or AI.

Tmkly•1h ago
Thanks. I think I have a good understanding of the business, having worked at the same place for 7 years. I am a leader in the sense of bringing technical ideas to the table, either for DX or product features. Often I use LLMs to sound out my ideas and thoughts and mould them into something I can go and talk about to higher-ups or other engineers to get buy-in. Also using LLMs a fair amount for writing tests which is great as I hate that bit!

Yeah I absolutely understand I'm in a better spot than juniors! But I worry slightly about the mid/longer term implications so want to make sure I'm prepared now rather than out of a job in some years due to AI.

I don't think I'm taking a moral stand against LLMs, but as you said, LLMs let complexity spiral and often I'm commenting on a PR for something a colleague has done which is clearly AI and not quite right. I try to care a lot about the code my team write but for sure we use LLMs.

Yeah you're right. I've always been front-end oriented, from when I was 15 making websites. But my fascination is more on the problem-solving side and coming up with new technical ideas, so I think if there were genuine problems to solve I could do that. I definitely see your point though.

Thanks for your views, very helpful.

Agile ain't as bad when done by clankers

https://github.com/sebs/rewelo
1•therebase•2m ago•1 comments

Paul R Ehrlich has died

https://www.nytimes.com/2026/03/15/books/paul-r-ehrlich-dead.html
1•LaserDiscMan•5m ago•0 comments

Construct of the Day

https://constructofthe.day
1•jasoncartwright•8m ago•0 comments

I Simulated 38,612 Countryle Games to Find the Best Strategy

https://stoffregen.io/posts/countryle/
1•st0ffregen•8m ago•0 comments

Show HN: AI agent autonomously used BoTorch to predict CFPB enforcement actions

https://github.com/sign-of-fourier/cfpb-complaint-enforcement
1•mshipman•10m ago•0 comments

Corruption erodes social trust more in democracies than in autocracies

https://www.frontiersin.org/journals/political-science/articles/10.3389/fpos.2026.1779810/full
2•PaulHoule•11m ago•0 comments

I Love FreeBSD

https://it-notes.dragas.net/2026/03/16/why-i-love-freebsd/
2•enz•13m ago•0 comments

West Sussex's Oracle rollout pushed back again as costs balloon 15 times

https://www.theregister.com/2026/03/16/west_sussex_oracle/
2•Brajeshwar•15m ago•0 comments

AI Unicorn ElevenLabs Is Making a $1B Effort to Restore 1M Voices

https://www.forbes.com/sites/zoyahasan/2026/03/13/ai-unicorn-elevenlabs-is-making-a-1-billion-eff...
1•stared•17m ago•0 comments

Operation Hormuz (1989)

https://www.lemon64.com/review/operation-hormuz/888
1•nurettin•18m ago•1 comments

Open source Chrome extension to customize Veesion alert notification sounds

https://github.com/MarcuC/custom-veesion-notifications-sound
1•molly997•19m ago•1 comments

Scientists discover a surprising way to quiet the anxious mind (2025)

https://www.sciencedaily.com/releases/2025/10/251027023816.htm
9•carlos-menezes•20m ago•3 comments

20 Years of Amazon S3

https://aws.amazon.com/blogs/aws/twenty-years-of-amazon-s3-and-building-whats-next/
3•easton•22m ago•0 comments

Meta up 3% in premarket as it plans mass layoff amidst increased AI spending

https://www.cnbc.com/2026/03/16/meta-ai-costs-mass-layoffs-20percent-up-premarket.html
3•rfarley04•22m ago•1 comments

Show HN: Merrilin.ai – Code blocks in your books, finally

https://tech.stonecharioteer.com/posts/2026/merrilin-code-blocks/
1•stonecharioteer•26m ago•0 comments

'Another internet is possible': Norway rails against 'enshittification'

https://www.theguardian.com/world/2026/mar/16/norway-rails-against-enshittifcation-deliberate-tec...
9•voxleone•26m ago•2 comments

ESM>CDN: A fast, Deno-friendly CDN for a world without 'NPM install

https://esm.sh/
3•kinderjaje•29m ago•1 comments

Perfex CRM: Unauthenticated RCE via PHP's S: deserialization format

https://nullcathedral.com/posts/2026-03-16-perfex-crm-unauthenticated-rce-insecure-deserialization/
1•nullcathedral•31m ago•1 comments

What Does Extreme Wealth Do to the Brain?

https://nymag.com/intelligencer/article/what-does-extreme-wealth-do-to-the-brain.html
1•cs702•32m ago•0 comments

Create a form in YAML, store it in GitHub, and share it as a live URL

https://docs.declarativeforms.com/getting-started/create-a-form
1•hirebarend•35m ago•0 comments

Dictare – Open-source voice layer for AI coding agents (100% local)

https://github.com/dragfly/dictare
1•dragfly•36m ago•0 comments

Promptfoo: Test your prompts, agents, and RAGs

https://github.com/promptfoo/promptfoo
2•ahamez•36m ago•0 comments

AWS S3 turns 20 and reaches 'exabytes'

https://www.theregister.com/2026/03/16/aws_s3_turns_20/
2•Brajeshwar•37m ago•0 comments

Dinit, the init system systemd should have been

https://unixdigest.com/articles/dinit-the-init-system-systemd-should-have-been.html
2•dither8•38m ago•0 comments

Node File Trace – determines which files a Node application needs to run

https://github.com/vercel/nft
1•javatuts•43m ago•0 comments

ROI Calculator for Field Service Businesses

https://fieldservicetools.com/tools/roi-calculator/
1•gemcoin•43m ago•0 comments

Badreads

https://www.funnest.world/badreads
2•surprisetalk•43m ago•0 comments

Three Modes of Cognition

https://kevinkelly.substack.com/p/three-modes-of-cognition
1•surprisetalk•43m ago•0 comments

A Zeckendorf Table

https://leancrew.com/all-this/2026/01/a-zeckendorf-table/
1•surprisetalk•43m ago•0 comments

Just Put It on a Map

https://progressandpoverty.substack.com/p/just-put-it-on-a-map
4•surprisetalk•43m ago•0 comments