The Balanced Engineer Newsletter

May 2026 • Issue # 45

ATTENTION: We are 1/3 of the way through the year. 33% already! Where did the time go?

These past two months have been a blur (oops I accidentally skipped a newsletter month again). I celebrated a wedding, did a cross-country trip with my three kids, attended Atmosphereconf AND Google Cloud Next, and had... let's just say a busy time at work. :)

Here are a few of the things I've enjoyed reading and listening to recently, as well as some of the content I've managed to make despite the insanity!

🔧 Technical Excellence

Building Effective Agents by Erik Schluntz and Barry Zhang (Anthropic)

Summary: This is Anthropic's foundational engineering post on what actually works when building agentic systems, drawing from a year of working with customer teams across industries. The authors draw a clear line between workflows (LLMs orchestrated through predefined code paths) and agents (LLMs dynamically directing their own work) and walk through five composable workflow patterns: prompt chaining, routing, parallelization, orchestrator-workers, and evaluator-optimizer, before getting into autonomous agents. The throughline is that the most successful implementations they've seen use simple, composable patterns rather than complex frameworks.

Why this resonates: I keep coming back to this one as I think more about the difference between "I've stitched together a workflow" and "I've actually built an agent." That distinction has gotten muddier in conversation as everything gets called an agent, and having the precise definitions to reach for has been useful. The advice to start with direct API calls before reaching for a framework also matches what I'm seeing in my own building, particularly as models get more capable and need less of a framework to operate.

🧠 Mental Models & Problem Solving

THE PEOPLE DO NOT YEARN FOR AUTOMATION by Nilay Patel (Decoder / The Verge)

Summary: Nilay introduces the term "software brain" to describe a way of seeing the world that fits everything into algorithms, databases, and loops. He argues this worldview has quietly run the business world for decades, but AI has supercharged it, and that's a big part of why public sentiment toward AI is so negative even as ChatGPT usage keeps climbing. He pulls in polling, including the finding that Gen Z, the generation using AI most, is also souring on it fastest, and makes the case that the gap between tech industry enthusiasm and regular people's experience isn't a marketing problem.

Why this resonates: As a software engineer I definitely have "Software brain". It's easy to forget that not everything benefits from being modeled as a database, and that the people I'm building for don't necessarily want their lives turned into a series of automated loops. It's a useful check on my own assumptions, especially when working on AI-adjacent products where the temptation is to assume more automation is always better.

⚖️ Work-Life Integration

I am in an abusive relationship with the technology industry from Salma Alam-Naylor

Summary: Salma writes about a moment of grief-induced burnout that started with a folk music video and turned into a reckoning with how AI has saturated every corner of her work. She talks about a friend who now feels like an administrator rather than an engineer because they only talk to Claude all day, the rise of company-wide AI mandates, and the strange pressure to adopt tools you don't actually enjoy or benefit from. It's a raw look at many of the drawbacks of the way things are going right now.

Why this resonates: I'm all in on AI tools but often feel myself thinking along these lines pretty frequently. It's hard to not despair at what has been lost and what has changed over time. I think that Salma really captures the mood many folks are feeling right now. Also, she recently made a music video and release a new song, and it is excellent!

What I've Been Building

A quick look at what I've been working on lately

Blog posts

• Embrace the Uncertainty
• Why I'm betting on ATProto (and why you should, too)

Overcommitted Podcast

• Ep. 49 | Using AI Agents in Software Development 2026: Current Uses and Future Possibilities

  A hosts-only episode where Bethany and I met up in person at the GitHub office before the Pragmatic Summit and talked about where AI agents actually are in early 2026 — what we're using them for, the async vs. sync mental model, custom agents and skills, parallelization with git worktrees, and a wishlist of where we'd like agents to go next (Google, please put AI in the calendar).

• Ep. 50 | Making Long Bets: How GitHub Next Shapes the Future of Software Engineering with Idan Gazit

  Idan Gazit, head of GitHub Next, joined Erika and me to talk about what it actually means to run an innovation team inside a big company. We got into the long-bets philosophy, why failing 80% of the time is a feature not a bug, the story of GitHub Blocks (a project he loved that didn't make it), and how the team's process has shifted post-Copilot toward holding onto prototypes longer rather than handing them off early.

• Ep. 51 | AI as a Power Up, Not Autopilot: Craig Dennis on Boosting Productivity and Software Development Education

  Craig Dennis is a developer educator at Cloudflare and the creator of AI Avenue, an educational docu-series. We talked about his journey from the Peace Corps to Treehouse to TwilioQuest to AI Avenue, why he believes "you have to touch the stuff" before you'll believe what AI can do, and the eye-opening conversations he's had with practitioners at companies like ElevenLabs and HeyGen about replacement, accessibility, and global reach.

• Ep. 52 | Overcommitted After One Year: Insights on Software Projects, Growth, and What's Next

  A hosts-only episode celebrating our one-year anniversary and our 52nd episode. Bethany, Erika, and I reflected on how the show has evolved from an accountability buddy group into a community of 160+ engineers, what's been harder than we expected (interviewing well, the admin grind), how we've each grown through it, and what we want to try next — including a panel format and more community learning sessions. We closed it out with our first-ever Overcommitted Superlatives.

• Ep. 53 | Scaling Search Engineering at DoorDash: From Monoliths to Custom Search Engines with Satish Saley

  Satish Saley spent years on DoorDash's search platform team and walked us through two big rebuilds: first replacing the monolith's search with a Kafka, Flink, and Elasticsearch pipeline, then later replacing Elasticsearch itself with a custom in-house engine built on Apache Lucene that cut latency in half and hardware costs by 75%. We dug into how to get leadership buy-in for a rebuild, the underrated art of migration work, and why he's stopped being so critical of legacy systems.

• Ep. 54 | Building Bulletproof Systems: Warren Parad on Software Engineering for High Availability

  Warren Parad is the CTO and co-founder of Authress and host of the Adventures in DevOps podcast. He joined Bethany and Erika to talk about how Authress stayed fully operational through the October 2025 AWS US-East-1 outage using DNS failover and multi-region strategies, what it actually takes to build a five-nines organization (hint: subtracting, not adding), and why hiring practices have to match the reliability culture you want. They wrapped up with a round of Never Have I Ever: SRE Edition.

• Ep. 55 | Building Your Own Tech Career Path - Bootcamp, Teaching & Big Platforms with Sabrina Goldfarb

  Bri Goldfarb is an engineer on the GitHub Copilot team and an instructor at Frontend Masters. She joined Erika and Bethany to talk about her non-traditional path into tech — video editor turned bootcamp grad — and how she's navigated chronic illness, burnout, and a career-defining stretch where she was working full-time at GitHub, running a startup, and teaching on the weekends all at once. A really honest conversation about pacing yourself, trusting the process, and not chasing prestige for its own sake.

• Ep. 56 | Run Toward Something - Career Growth, Mentorship & Work-Life Balance with Dave Schwantes

  Dave Schwantes is a senior software engineer at GitHub who I co-lead the engineering mentorship program with. We talked about why he runs toward meaningful work rather than away from burnout, how mentorship became his primary form of engineering leverage (including the in-house bootcamp he helped build at Instacart), and how his definition of career happiness has evolved across different life stages. Also: he's building a messaging app for his kid and trying to scare squirrels off his bird feeder with metal music.

• Ep. 57 | Build Real Tools, Skip LeetCode: Systems Programming for Career Growth with John Crickett

  John Crickett is a 30+ year engineer and the creator of Coding Challenges, a newsletter with 90K+ subscribers that gives engineers real-world projects to build instead of LeetCode problems — your own Redis, your own Git, your own shell. We talked about how Coding Challenges went from a $17 domain to 1,500 signups in a single weekend, why he thinks a lot of fundamental knowledge has been quietly eroding for decades (long before AI), and what he sees in fields most of us forget exist, like the still-growing mainframe market. We wrapped up with a quiz guessing whether obscure programming languages are real or made up.

Join us on Discord!

Our next book club for The Staff Engineer's Path starts soon!!

Come hang out with the Overcommitted community on Discord! We're chatting about engineering, career growth, and all the things we're learning together.

Have comments or questions about this newsletter? Or just want to be internet friends? Send me an email at brittany@balancedengineer.com or reach out on Bluesky or LinkedIn. Until next month!