What are the key takeaways from this The Bike Shed episode?

Key insights include: **AI-Assisted Onboarding:** Use Claude in plan mode to identify which existing specs to examine first when starting new projects, rather than having it write code directly. Ask the AI to flip the script and question your feature approach, then distill those questions for stakeholders across time zones, reducing wait time from hours to minutes while maintaining human validation.; **Consultants as Question Askers:** Professional consulting centers on asking the right questions early, especially for new features that establish future patterns. Use AI to surface edge cases and considerations not in acceptance criteria, then validate with product stakeholders. Understanding what you don't know matters more than demonstrating existing knowledge, particularly when onboarding to unfamiliar domains.; **Suspenders Architecture Evolution:** The gem shifted from inheriting Rails generator classes to becoming a simple wrapper for Rails new that passes three options: Postgres database, skip test framework, and an application template file. This rewrite removed 5,000 lines of code, added 1,000, and reduced generation time to eight seconds while making contributions significantly easier.

What did Steve Polito discuss on The Bike Shed?

Steve Polito discusses the major rewrite of Suspenders, Thoughtbot's Rails application generator gem that existed since 2008. The conversation covers using AI for onboarding, the technical architecture shift from Rails engine to application template, and how the gem packages Thoughtbot's Rails conventions including RSpec, FactoryBot, Heroku deployment, and strong migrations into a single command. Key topics include: **AI-Assisted Onboarding:** Use Claude in plan mode to identify which existing specs to examine first when starting new projects, rather than having it write code directly. Ask the AI to flip the script and question your feature approach, then distill those questions for stakeholders across time zones, reducing wait time from hours to minutes while maintaining human validation.; **Consultants as Question Askers:** Professional consulting centers on asking the right questions early, especially for new features that establish future patterns. Use AI to surface edge cases and considerations not in acceptance criteria, then validate with product stakeholders. Understanding what you don't know matters more than demonstrating existing knowledge, particularly when onboarding to unfamiliar domains..

How long is this episode of The Bike Shed?

This episode is 41 minutes long. SignalCast provides an AI-generated summary so you can get the key insights in about 3 minutes.

The Bike Shed

493: Suspenders with Steve Polito

February 10, 2026

41 min episode · 2 min read

Steve Polito

Episode

41 min

Read time

2 min

Topics

Artificial Intelligence, Software Development, Product & Tech Trends

AI-Generated Summary

Published Feb 11, 2026

Key Takeaways

✓AI-Assisted Onboarding: Use Claude in plan mode to identify which existing specs to examine first when starting new projects, rather than having it write code directly. Ask the AI to flip the script and question your feature approach, then distill those questions for stakeholders across time zones, reducing wait time from hours to minutes while maintaining human validation.
✓Consultants as Question Askers: Professional consulting centers on asking the right questions early, especially for new features that establish future patterns. Use AI to surface edge cases and considerations not in acceptance criteria, then validate with product stakeholders. Understanding what you don't know matters more than demonstrating existing knowledge, particularly when onboarding to unfamiliar domains.
✓Suspenders Architecture Evolution: The gem shifted from inheriting Rails generator classes to becoming a simple wrapper for Rails new that passes three options: Postgres database, skip test framework, and an application template file. This rewrite removed 5,000 lines of code, added 1,000, and reduced generation time to eight seconds while making contributions significantly easier.
✓CI Testing Generated Apps: Configure continuous integration to actually run suspenders new, generate an app, scaffold a resource, and execute the test suite inside the generated application. This prevents the common failure mode where the gem's tests pass but apps generated by the tool have broken CI, ensuring end users receive working applications from day zero.
✓Two-Commit Audit Trail: Generate an initial commit containing everything Rails new produces, followed by a second commit showing only suspenders modifications. This creates a clear diff for teams to audit what the gem adds, enables easy customization of specific settings, and provides valuable context for future git blame investigations when questioning configuration choices months later.

What It Covers

Steve Polito discusses the major rewrite of Suspenders, Thoughtbot's Rails application generator gem that existed since 2008. The conversation covers using AI for onboarding, the technical architecture shift from Rails engine to application template, and how the gem packages Thoughtbot's Rails conventions including RSpec, FactoryBot, Heroku deployment, and strong migrations into a single command.

Key Questions Answered

•AI-Assisted Onboarding: Use Claude in plan mode to identify which existing specs to examine first when starting new projects, rather than having it write code directly. Ask the AI to flip the script and question your feature approach, then distill those questions for stakeholders across time zones, reducing wait time from hours to minutes while maintaining human validation.
•Consultants as Question Askers: Professional consulting centers on asking the right questions early, especially for new features that establish future patterns. Use AI to surface edge cases and considerations not in acceptance criteria, then validate with product stakeholders. Understanding what you don't know matters more than demonstrating existing knowledge, particularly when onboarding to unfamiliar domains.
•Suspenders Architecture Evolution: The gem shifted from inheriting Rails generator classes to becoming a simple wrapper for Rails new that passes three options: Postgres database, skip test framework, and an application template file. This rewrite removed 5,000 lines of code, added 1,000, and reduced generation time to eight seconds while making contributions significantly easier.
•CI Testing Generated Apps: Configure continuous integration to actually run suspenders new, generate an app, scaffold a resource, and execute the test suite inside the generated application. This prevents the common failure mode where the gem's tests pass but apps generated by the tool have broken CI, ensuring end users receive working applications from day zero.
•Two-Commit Audit Trail: Generate an initial commit containing everything Rails new produces, followed by a second commit showing only suspenders modifications. This creates a clear diff for teams to audit what the gem adds, enables easy customization of specific settings, and provides valuable context for future git blame investigations when questioning configuration choices months later.

Notable Moment

The discussion reveals how thinking about prose and arguments as dependency graphs enables structural editing. Each premise becomes a node with dependencies, and valid writing paths must follow topological sort order where you only visit nodes after their dependencies. This separates edits that change the graph structure from those that merely reorder the traversal through unchanged dependencies.

Know someone who'd find this useful?

Episode Transcript

Hello, and welcome to another episode of the Bike Shed, a weekly podcast from your friends at Thoughtbot about developing great software. I'm Joelle Kenville. And today, I'm joined by fellow Thoughtbotter, Steve Pulido. Hey, Joelle. Hey, everyone. Glad to be back. Together, we're here to share a bit of what we've learned along the way. So, Steve, what's new in your world? Well, aside from it being, like, two degrees every day for the past week, I started a new new client project, which is always exciting. So, yeah, that's been good. And I've been using agentic AI to kinda help me onboard a bit. So that's been that's been fun. I think this is the first time I've, like, really leaned into that. In what ways do you use agentic AI for for onboarding? Is this, like, explain to me how this module work? Is this, generate diagrams for me about the architecture? Tell me a little bit more about your your onboarding process using AI. It's, like, not experimental, but, like, I don't know what my, I'm honing in on it sort of. Right? So, like, each new project I roll on, I'm not, like, necessarily consistent with how I use it. But in this one, it has been primarily, like one thing I've consistently been doing is, like, I look at my ticket, get a sense of, like, what it's asking me to do, and then I'll I'm using cloud code. Right? So I'll start a new chat with cloud usually in plan mode because that's what I've been kinda I'm finding a lot of use and novelty in that at the moment versus just having it, like, write the code. So thing I've been doing consistently is just saying, hey. Here's the task at hand. Can you tell me which existing spec to look at first? Because I I still find a lot of value in TDD, and I I just wanna get to that point as quickly as possible. And with a new project, you know, it's a little hard to know where exactly that is because every project is a little different. So that's been helpful. That's one thing I've been doing a lot. And then also there's benefits because we're across time zones, so the client is three hours behind me. So, you know, I can't necessarily wait three hours to get an answer on something. So Claude is helpful there too. So it kinda gives me this, like, buffer, I guess, where I I have this back and forth conversation with it. It's not so much writing code for me, but it's just, like, pointing me in the right direction, giving me some artifacts to work off of. Then we have a bit of a conversation where it it flips the script, and it'll ask me how I envision the feature. And sometimes I can answer that. And other times, I'm like, I don't know. I'm new to this project. But …

Get the full transcript (7,240 words) + summary by email — free

One-time email with the complete transcript and AI summary of this episode. No account needed.

One email, no spam. We’ll also show you what SignalCast does.

Browse all The Bike Shed transcripts →

You just read a 3-minute summary of a 38-minute episode.

Get The Bike Shed summarized like this every Monday — plus up to 2 more podcasts, free.

Pick Your Podcasts — Free

#879: How to Reinvent Yourself and Make Bold Moves — Decisive Stories from Matthew McConaughey, Steve Young, Samin Nosrat, and More

Aug 12

Books, tools, and gear mentioned in this episode

SignalCast may earn commission on purchases via these links.

Tools

FactoryBot
“the gem packages Thoughtbot's Rails conventions including RSpec, FactoryBot, Heroku deployment, and strong migrations into a single command.”
SuspendersBy guest
by Thoughtbot
“Steve Polito discusses the major rewrite of Suspenders, Thoughtbot's Rails application generator gem that existed since 2008.”
Rails
“The gem shifted from inheriting Rails generator classes to becoming a simple wrapper for Rails new that passes three options: Postgres database, skip test framework, and an application template file.”
Heroku
by Salesforce
“the gem packages Thoughtbot's Rails conventions including RSpec, FactoryBot, Heroku deployment, and strong migrations into a single command.”
Claude
by Anthropic
“Use Claude in plan mode to identify which existing specs to examine first when starting new projects, rather than having it write code directly.”
RSpec
“the gem packages Thoughtbot's Rails conventions including RSpec, FactoryBot, Heroku deployment, and strong migrations into a single command.”