Have you noticed faster pace of development?
Have you seen changes to code quality or code review?
Do teammates that use these tools complete sprint tasks faster than those who don't?
My team has adopted Claude Code extensively, and the number of daily and weekly PRs we have closed has increased significantly. I’ve noticed that we’re also more willing to commit to more projects. My team benefits from most of our code being written in TypeScript. However, some other teams with legacy code bases seem to have a bit of a harder time using these tools compared to us.
One thing that surprises me with a AI is you can have people working on the same code base. Some can be very effective with AI, yet there are others working on the exact same code base who cannot get good results. Some people don’t really seem to be taking the time to get good at writing prompts and plans before having agents execute on them.
My employer has a slack channel where we can share prompts, tips about how to use AI outside of coding, Q&A. I've learned a lot from my coworkers there.
> Have you noticed faster pace of development?
Yes, our org has had a 50% increase in PRs since Opus 4.5 released.
> Have you seen changes to code quality or code review?
Yes, significantly more bugs (no exact number), but consider it maybe 3-4x in volume. However, nothing catastrophic and everyone just uses AI for fast-follow fixes anyways. The company as a whole is embracing this style of development for better or worse.
> Do teammates that use these tools complete sprint tasks faster than those who don't?
Yes, but my entire team uses them. I’d say the ones who use it more effectively (crazy skill setups, better tooling/commands, better scaffolding) finish much faster. Probably 80% of my team still uses Cursor in the one-shot way with very vague requirements, and don’t have the AI connected to github, jira, slack, etc which can actually feed really important context into decision making.
If I do something more than once a day, I write a custom slash command for it. This has personally 2x’d my pace.
using ai coding assistants has allowed me to onboard faster because i can take a new ticket and ask the agent to help orient me in the codebase about where to start looking.
BUT the downside is that i'm much slower at building a cohesive mental model of the codebase. this has two problems:
- it encourages continued dependence on the ai tools because its hard to fall back on my own knowledge
- imposter syndrome is magnified 100x
People can build faster, but choosing what to build and maintaining consistency across tools is becoming harder.
The coordination problem feels bigger than the coding problem now.
> Have you noticed faster pace of development?
Yes, and it terrifies me.
> Have you seen changes to code quality or code review?
Yes, because of the volume of PRs, I am sure some people are just approving changes with a quick glance
> Do teammates that use these tools complete sprint tasks faster than those who don't?
Best engineers got better
What terrifies me more like cultural thing:
1. Engineers who push a lot of code are trying to get promoted faster so they won't feel the pain
2. Engineers who genuinely care and trying to land high quality work are getting labeled as "progress blockers" and getting negative outlook, because they won't be promoted, they need to support the slop added by other engineers. And I don't like this state of the culture.
Also engineers who want to get promoted quickly, usually they are very good at talking and can present a tiny skill template as something huge developed over 2 quarters of experience with latest and greatest, while people who actually do the work are more silent and heads down executing, I want them to be promoted but can't impact a lot of things
For a time, its all about shipping faster. And rushing things out.
Then there's a major bug/incident/unhappy customer, and management swings back to code quality
AI coding is definitely speeding me up, but not quite in the magic way people at smaller companies might be used to.
As a good code complete it's awesome. As a brain storm buddy, very good. AI code reviews are awesome and very low risk - if it makes a mistake it's usually just being too cautious.
As an end to end agentic coder -pretty bad track record. Giving it smaller tasks? Pretty good, saving me many hours of boilerplate/cruft a week at least.