Infinity Interactive

02/26/2026

Our VP of Technology made biang biang noodles from scratch while an AI agent built 13 story points across 3 pull requests. His active involvement for that stretch was about ten minutes.

The noodles aren't the point, though. The 105 minutes of architecture planning that happened before any code existed — that's the point.

Over the past year, we've been developing a four-phase methodology for AI-augmented development across our client projects: Plan → Assign → Build → Retro. Eric ran the full workflow as a training exercise for the team, documented everything obsessively, and wrote up what he learned.

The post covers how detailed planning compresses multi-month timelines, the three-layer automated review system that caught a timezone bug a human would have shipped, and an honest accounting of what he'd do differently next time.

It's a long read. It's worth it if your team is experimenting with AI development tools and finding the results inconsistent.

Link in the comments!

02/04/2026

Three days before the end of a five-week sprint, our client's VP of Marketing joined a status call. She took one look at our nearly-complete portal and spotted the problem immediately: we'd been building from outdated brand guidelines.

The documentation we received at kickoff was from a previous brand iteration. Nobody realized it until that moment.

So: 72 hours left, and we needed to rebuild the navigation, hero, footer, and landing page patterns from scratch. We shipped on time, everything passing QA.

The difference between "three hard days" and "three extra weeks" came down to infrastructure we'd built from sprint one, not because we expected trouble, but because that's just how we work.

Link to our new post on the ii notebook in the comments!

01/15/2026

Talk to developers about AI and you’ll get one of two reactions: breathless enthusiasm or principled rejection.

We think both miss the mark.

After a year of integrating AI tools into our workflows and teaching clients to do the same, we’ve noticed two failure modes. Abdication: letting AI make decisions that require human judgment, shipping code nobody understands. Avoidance: refusing to engage with useful tools because the hype is so off-putting.

Between those extremes is what we call assistive AI. The tools handle mechanical friction. Humans keep ownership of architecture, problem-solving, and the decisions that actually require experience.

We wrote about how we think about this, what it looks like in practice, and what to ask when you’re evaluating teams that claim AI makes them faster.

Link in comments.

01/06/2026

Every developer knows the Monday morning standup scramble. What did I work on Friday? Which tickets did I close? What's still in progress?

Our VP of Technology Eric Wagoner got tired of the daily archaeology through commit logs and Jira tickets, so he built an n8n workflow that gathers his GitHub and Jira activity automatically and generates a natural, conversational standup update delivered to his inbox every morning at 8 AM.

He expected the time and focus savings. The daily morale boost was a pleasant surprise.

Eric's sharing the complete workflow JSON so you can set it up yourself. One config node holds all your project-specific values, so spinning up a new copy for each project takes minutes. When the engagement ends, just delete the workflow.

Link in comments!

12/23/2025

When you need more technical capacity, your options aren't great:

→ Hire full-time (expensive, slow, doesn't flex)
→ Stretch your existing team (risky, limited)
→ Hire freelancers (flexible but no continuity)

Each solves one problem while creating others.

What if there was a model that worked more like your office's lawn care service? A team that just shows up with the right people and the right skills, scales up or down as needed, and handles it.

That's what we built with F**T.

New post: What Your Lawn Care Company Can Teach You About Hiring Developers

12/16/2025

"I refreshed, but nothing changed."

Every developer has heard this. But when you're building a Flutter PWA to take advantage of ease of distribution and updates, yet your QA team reports not being able to see the new fixes and features, you know something funny is going on. And when you see this behavior across multiple releases over two months, and hard refreshes don't work, and restarting the browser doesn't work, and iOS Safari is fine but macOS Safari isn't, you know you're in for a journey.

Development went smoothly. Weekly releases, rapid iteration, happy client. Then Safari started holding onto cached content like it was the last lifeboat on the Titanic.

What followed was a two-month deep dive into service worker internals, custom update mechanisms, git hash version checking, CORS surprises, and one humbling false alarm where the "bug" turned out to be our system working exactly as designed.

The good news: our QA process caught every issue before production. Not a single user was affected. The lessons learned (about Safari's unique caching behavior, about building update mechanisms from day one, about keeping service workers focused on static assets) will inform every PWA we build going forward.

We wrote it all up so you don't have to learn it the hard way.

Link in comments!

12/09/2025

Documentation drift is one of those problems every development team knows too well. Setup instructions that reference config files that were replaced months ago. API docs with sample code that no longer compiles. New team members spending hours troubleshooting issues that an updated doc could have prevented in five minutes.

Nobody lets docs rot on purpose. But when you're focused on shipping features and solving client problems, documentation updates always seem to get pushed to next sprint.

Our VP of Technology Eric Wagoner built Diderot to solve this. It's a Claude Code sub-agent (named after the philosopher who edited the Encyclopédie) that analyzes what's changed in a codebase and updates the documentation to match. What used to take 2-3 hours of git log archaeology every couple weeks now takes 15 minutes.

This is how we think about AI at Infinity Interactive. Not as a replacement for human judgment, but as a way to amplify it. Diderot doesn't decide what's worth documenting or how to structure information for users. Those are human calls that require experience and taste. He handles the mechanical synchronization so our developers can focus on architecture, problem-solving, and the work that actually benefits from their expertise.

Eric built a similar tool called Ray for extracting design tokens. Same philosophy: AI handles the tedious transcription work so humans can stay focused on the decisions that matter.

We've always been known for the quality of our work. AI doesn't change that standard. It just means we're not spending hours on tasks that pull us away from delivering it.

More in the comments!

11/22/2025

We rebuilt our entire website in 30 days.

Here's why, and how we did it with AI as our assistant, not our architect.

Why We Rebuilt
============
When we set out to rebuild Infinity Interactive's website, we weren't just updating design. We needed to tell a different story.

Our business had evolved. We were no longer just technical people helping technical people. We'd become the team that small business owners and directors call when they need technical problems solved (but don't want to manage the technical details themselves).

Our legacy site didn't reflect who we are today or how we work. We needed to show prospects that we:

• Take technical headaches completely off their hands
• Deliver quickly without compromising quality
• Own every line of code we write

The Human-First Approach
====================
As experienced developers who use AI tools in our daily work, we leveraged Claude Code and Cursor IDE throughout this project. But not the way you might think.

We didn't let AI make our decisions or write our software. Instead, we used it the way seasoned professionals use any powerful tool: to accelerate the mechanical parts while we focused on architecture, problem-solving, and quality.

We owned:
All architectural decisions, design system, code review, business logic, problem-solving

AI helped:
Mechanical refactoring, boilerplate generation, documentation formatting, content migration patterns

The Result
========
Built on Next.js 15 with static site generation:
• 200+ static pages
• ~70 migrated blog posts
• 5 case studies
• 129 SEO-preserving redirects

All delivered by 3 developers in about 30 days of work time (while handling client projects). Every line of code was planned, reviewed, and owned by humans. Claude Code accelerated our work; it didn't replace our expertise.

Why This Matters
=============
The distinction matters. Our clients hire us for our judgment and ability to solve complex problems. AI tools make us more efficient, but they don't replace the years of experience we bring to every project.

This is how we approach all our work: Human expertise first, smart tools second.

Read the full story about our process, lessons learned, and how we balanced speed with quality:

How three experienced developers built a modern marketing website with 200+ static pages in 30 days of work time using Next.js 15, Tailwind CSS v4, Claude Code, and Cursor IDE—without compromising human ownership of every line of code.

11/08/2022

Happy Tuesday and welcome to the latest Infinity interview! Get to know our developer Jeremy Tarver as we talk Toli, treehouses, and trips down lazy rivers.

An interview with Jeremy Tarver

08/05/2022

Happy Friday and welcome to the latest installment of our employee interview series! This week, meet Suzanne Raphael, whose intrepid, adventuring spirit, mixed with a hearty spoonful of overthinking, makes her Infinity’s perfect QA evangelist.

An interview with Suzanne Raphael

06/17/2022

Three cheers for Friday and for our latest Infinity interviewee! Meet Matt Holtz, an Infinity consultant and brand new dad. We gab about his golf game, his new baby girl, and his great big dog.

https://iinteractive.com/notebook/2022/06/17/meet-matt-holtz/