AI Engineering Engagement @sonarsource. Exploring the AI coding quality challenge!
Stroud UK
Joined July 2009
- Tweets2.3K
- Following722
- Followers907
- Likes975
Pinned Tweet
Present a good team with problems & they'll thrive on finding solutions. Ask them to implement your solution & they'll only find problems.
Here’s my deep dive on my conversation with Simon Vane, Staff Engineer at M-KOPA, I look at how review, tests and design conversations fit together when agents deliver faster than people can inspect.
Read the full deep dive on Substack: thenextcommit.substack.com/p…
It's tempting to put all the time agents save into building more features. But without strong foundations, adding more complexity will make every new feature harder to add.
Read my newsletter for more thoughts: thenextcommit.substack.com/p…
Which parts of pairing, TDD and quality still matter when agents write the code?
I asked my old friend Simon Vane from MKopa on to The Next Commit, and we worked through what can move into agent instructions and automated checks, and what still needs people.
Listen: thenextcommit.substack.com/p…
How can agents have enough freedom to help, without removing the safeguards that make their work trustworthy?
Find out in the deep dive on my conversation with Shashank Patwardhan, Application Security Lead at @ifxpayments.
Read: thenextcommit.substack.com/p…
Tom Howlett retweeted
AI agents know architectural patterns, but they still need boundaries.
@codeliftsleep shares his repeatable process for setting constraints that keep agents aligned with human architectural decisions.
Watch his discussion on The Next Commit with @leantomato: hubs.la/Q04BMK520
Check out Matt's book, Grokking Software Architecture: hubs.la/Q04BMK8r0
foundations remain as important today as they did in August.
continue to invest accordingly :)
organic architecture
good codebases have always had strong foundations and constraints. these files go here. this kind of code goes there. the codebase is held together either by catering to the lowest common denominator developer using conventional frameworks and languages, or by strict adherence to a style guide enforced by code review.
the arrival of an infinite number of monkeys armed with keyboards have obsoleted the latter. codebases designed for agents need hard constraints so that even the dumbest agent (and pilot) can contribute meaningfully. the chaos must be tamed.
refactoring your codebase and choosing a tech stack with strong type systems/compiler diagnostics has far greater impact than AGENTS.md and tweaking your skills. hard constraints and well thought out architecture guide agents to do the right thing by default. the best rules are in your code.
invest accordingly.
Where should a coding agent's freedom stop, and who decides?
I talked it through with Shashank Patwardhan, Application Security Lead at @ifxpayments , this week on The Next Commit.
We cover agent access to human judgment and accountability, here: thenextcommit.substack.com/p…
What would you need to put in place before you trust your agent with your architecture?
We dig into it with Matt Erman @codeliftsleep, Senior Software Engineer at @Blackboard, on The Next Commit, and the deep dive breaks down how he builds that confidence from evidence, small steps and automated checks.
Read the deep dive: thenextcommit.substack.com/p…
Developers fear what agents will do to their architecture, but agents can be great architects if you give them constraints to work within! 🎙️
My guest on The Next Commit this week is @codeliftsleep. Matt Erman is the author of Grokking Software Architecture. He shares some wise words on how we can maintain good architecture in the age of agents.
Listen: thenextcommit.substack.com/p…
How do your teams stay aligned at agentic speed in a highly competitive industry?
substack.com/home/post/p-218…
Time for a Next Commit deep dive into how @SonarSource's very own Quentin Chevrin uses agents to develop the billing system for SonarQube.
It includes a great clip from the podcast in which Quentin explains how he scopes his prompts and prepares his commits and PRs for review.
thenextcommit.substack.com/p…
Tom Howlett retweeted
Episode 4 of @TheNextCommit, out now!
@leantomato talks to Quentin Chevrin, one of our front-end developers.
They cover handing implementation to a coding agent, the work he still doesn't delegate, & why code review is now the slowest part of shipping👇
thenextcommit.substack.com/p…
I’m excited to share this week’s The Next Commit episode where I interviewed @SonarSource's very own Quentin Chevrin. Quentin has taken a very thoughtful and considered approach to adopting agents and I loved this very honest conversation. Have a listen: thenextcommit.substack.com/p…
Not a podcast person? Prefer a long read, or asking Claude or GPT to read it for you?
I’ve written a deep dive into this week’s The Next Commit episode with Michael Tweed. We explore how Skyscanner turns local AI experiments into shared engineering practice, and how engineers’ work shifts toward guiding agents and verifying what they produce.
thenextcommit.substack.com/p…
Are your teams sharing how they work with agents?
Today on The Next Commit, Michael Tweed explains how @Skyscanner turns local experimentation into shared skills, infrastructure and standards without imposing one workflow on every engineer.
thenextcommit.substack.com/p…
One run won’t tell you how coding agents work. You need repeated experiments and traces to see the patterns. I burnt some tokens benchmarking agents and reasoning settings with Sol. Codex and I wrote up what changed when I turned reasoning effort up and down: thenextcommit.substack.com/p…
If you’d prefer a long read to this week’s @TheNextCommit podcast, here’s my analysis of the conversation with @tottinge.
Agents don’t make XP obsolete. They give testing, refactoring and evolutionary design new leverage, while making human judgement matter more.
thenextcommit.substack.com/p…
With the right guidance, agents can write quality code at extraordinary speed.
Tim Ottinger (@tottinge) gives them room to work, but there’s one situation where he always tells them to stop and collaborate.
Can you guess when?
thenextcommit.substack.com/p…