Hey, friends. Let’s break it down.
A ton of smart people around me still treat AI-assisted software development as a fad, some sparkly, overhyped gimmick. I used to let it slide, thinking, they’re smarter than me, right?
Here’s what really matters: dismissing LLM‑driven development today is like ignoring electricity in the 1880s. LLMs may not be perfect, but if you’re still betting against them in code, you’re betting on sand.
Getting Real on AI Agents
If you’re just pasting ChatGPT code snippets into your editor and calling it AI work, you’re doing it wrong.
Real practitioners are using LLM agents. These agents:
- Navigate your repo autonomously, run tests, inspect files, lint, compile, even Git commit
- Make real toolcalls you define
- Handle orchestration like a Makefile plus intuition
You could spin up a basic coding agent over a weekend. The sophistication comes from the tooling and how you wire it together.
If you’re still guilting AI for being hallucination-prone, here’s the thing: developers build guardrails. Agents compile, run checks, and refine until green. Hallucination stops mattering unless you ignore the logs.
So when people moan about AI writing “junior-level” code, they miss the point. An LLM supplement is exactly that, a supplement. To scale development across repetitive tasks, onboarding drives, scaffold factories, yeah, it’s worth it.
Your Job Didn’t Disappear
You’re still responsible for code you merge. Every integration. Every commit goes through you. If the AI-generated stuff sucks, you guide it, refine it, still your call.
You know deep down: repetitive code is painful. Writing tests, scaffolding, boilerplate, an AI agent eats that stuff for breakfast. You get the dopamine of logic working, then you jump in, tweak, polish, own it. That’s still real code work.
But Some Folks Refuse to Read
There’s a cultural irony in the dev world. Intellectual property policing is intense, but the same devs glance over license violations when it’s convenient. Then they freak over possible reposted code from LLMs. It’s projection. If you can’t handle reading output, maybe real-world teamwork isn’t for you.
Rust, Golang, and the Language Bias
Some team members say “AI fails at Rust” or fancy-pants languages. Cool. So don’t use AI for those parts. Use it everywhere else.
We build mostly in Go and Python at e10 and frankly, agents crush with those. If your language choice is tripping AI up, acknowledge it and plan around it.
Here’s the Bottom Line
LLM‑assisted code isn’t clever magic; it’s reliable machinery. You still curate. You still lead. You still understand the logs and the stack traces.
The only real question: when did you stop trusting your own skill?
What this means for e10:
We’re launching internal agent pipelines. Starting basic, scaffold handling, dependency updates, test bootstrapping, but the plan is clear: reduce grunt work, level up human creative effort. If you’re on the team, get familiar. If you’re outside looking in, adjust your lens.
AI coding isn’t optional anymore. It’s table stakes. Ignore it, and yes, you’re all nuts.