Richard Seroter's Architecture Musings

19 Aug 2026

Today's links look at doc-driven development, why harnesses are basically situated agents, what's new in Go 1.27.

18 Aug 2026

Today's links look at data about AI usage patterns in software teams, some best practices for AI code reviews, and what to do when the hard part stops being hard.

17 Aug 2026

Today's links look at how companies are doing agentic engineering in real life, what to do about skill sprawl, and whether you need skills at all.

14 Aug 2026

Today's links look at how to speed up generative UI, which patterns and problems in multiagent systems to be aware of, and whether software factories actually work.

13 Aug 2026

Today's links look at the flashy Gemini 3.7 Flash model, why reviewing code is a skill, and what technical debt in the AI era looks like.

13 Aug 2026

Today's links look at why local models won't win, whether TDD inside the agent loop makes sense, and where a watermark can hide in plain text.

11 Aug 2026

Today's links look at whether agents are killing the UI, whether agents can use computers yet, and a fresh definition of the product role.

10 Aug 2026

Today's links look at why 'code was never the hard part' is insulting, what code review are for, and why open source matters for AI.

7 Aug 2026

Today's links look at when a loop should stop, how WebMCP is picking up momentum, and what has to be true to rewrite the code all the time.

6 Aug 2026

Today's links look at the new Agent Plugins spec, when you should self-host inference, and what nobody tells you about writing agent skills.

5 Aug 2026

Today's links look at a big change in Google AI leadership, what name we should use when referring to AI coding, and whether lacking a long-term career plan is fine.

5 Aug 2026

Today's links look at how to work with 5 coding agents simultaneously, scaling agents with session-aware load balancing, and why engineering management is a career change instead of a promotion.

4 Aug 2026

Today's links look at the economic benefits of refactoring, the one line that 95% of agent skills are missing, and how to give (and take) credit in big tech companies.

30 Jul 2026

Today's links look at what a loop is, how to apply eval-driven development, and why GitHub might be the wrong shape for this AI world.

30 Jul 2026

If you’ve been a software developer for more than five minutes, then you probably have some old code running somewhere. It probably works fine, so who cares if it’s a little dusty and based on an

29 Jul 2026

Today's links look at how much you can delegate to agents, how we killed reading and can get it back, and why your moat shouldn't be the models.

28 Jul 2026

Today's links look at a big update to the Model Context Protocol, why it's good to be bored, and how to vibe code your first Android app.

28 Jul 2026

Today's links look at the AI productivity paradox, why mentoring matters more in the AI era, and why you should stop engineering your agent harness.

24 Jul 2026

Today's links look at what the right amount of spec is for agentic development, what data you need to determine the ROI of AI, and which architecture mistakes make systems hard to change.

23 Jul 2026

Today's links look at where the real AI competition is at, the challenge of measuring AI ROI, and why you should do some pre-factoring in your codebase.

22 Jul 2026

Today's links look at the open source agent toolkit in 2026, why AI keeps pushing us to ship, and how to tell if an engineer using AI is growing.

22 Jul 2026

Maybe it’s just me, but it feels like most of the agentic coding tools out there are single-player. An individual developer engages in a coding session with one or more agents, and the non-code

21 Jul 2026

Today's links look at the latest Google Gemini models, why reliability is a discipline and not a feature, and how Netflix serves LLMs in-house.

21 Jul 2026

Today's links look at how to earn taste and judgement, why loop engineering is mostly just broken SDLC wearing a costume, and how to design APIs for agents.

18 Jul 2026

Today's links look at what ten AI film crews teach us about agent teamwork, how to cut token usage by 94%, and why model routing is simple until it isn't.

16 Jul 2026

Today's links look at why the prototype is a question and not a product, five trends that define modern AI engineering, and why "playing politics" isn't a bad thing.

15 Jul 2026

Today's links look at how Netflix built a real-time look at their service topology, why you should control the ideas and not the code, and why 90% done rarely means that you're almost done.

14 Jul 2026

Today's links look at challenges with persistence for long-running agents, why we keep arguing against Jevon's paradox, and securing the AI supply chain.

13 Jul 2026

Today's links look at how to stop being the code review bottleneck, why you should still write code in 2026, and why TypeScript 7.0 was rewritten in Go.

11 Jul 2026

Today's links look at why you need to own the outer loop, why AI coding agents still need clear specs, and why multi-tenancy isn't all about databases.

9 Jul 2026

Today's links look at how AI is changing what employers want from new hires, which new models are getting attention, and why you should do hard (meaningful) things.

8 Jul 2026

Today's links look at why AI mandates can be good, why your configuration-as-code approach is a liability, and how AI can fix traffic congestion.

8 Jul 2026

Today's links look at how to help workers reskill, why you should write code instead of specs, and whether the PRD is dead.

6 Jul 2026

Today's links look at traits that set the best leaders apart, ten moments that defined AI's first half, and why we're seeing the twilight of the chatbot.

2 Jul 2026

Today's links look at why experienced (engineering) managers get stuck, which vulnerabilities are most risky for MCP servers, and the do's/don'ts for LinkedIn posts.

1 Jul 2026

Today's links look at tools to bolster your AI coding stack, driving the agent quality flywheel from your coding agents, and why AI should shrink teams and not headcount.

30 Jun 2026

Today's links look at how enterprises see the future of (AI) engineering, whether markdown is the application layer of AI apps, and how to think about agent memory.

29 Jun 2026

Today's links look at essential testing patterns for AI agent development, why AI adoption is overloading middle managers, and the problem with prompt debt.

28 Jun 2026

That was fast. We’ve moved from prompting an LLM, to providing instructions to an agent, to having agents prompting other agents in a loop. “Loop engineering” is all the rage among the AI elite

26 Jun 2026

Today's links look at loops versus prompts, where the agent harness becomes tech debt, and how Gemini can help you avoid jetlag.

26 Jun 2026

Today's links look at ping-pong pairing with AI, why configurations belong in a database, why we shouldn't be using agents to buy the same answer twice.

24 Jun 2026

Today's links look at why the mom-and-pop SaaS era has arrived, why Europe cloud sovereignty push may backfire, and where the AI productivity bill comes due.

23 Jun 2026

Today's links look at why old software was so fast, how to become ridiculously good at Kubernetes, and how to secure the future of AI agents.

22 Jun 2026

Today's links look at the case against building your own agent platform, what autonomous long-running code agents are all about, and forces impacting the transition from manager to leader.

19 Jun 2026

Today's links look at why AI coding isn't working for your team, the death and rebirth of programming, and the keys to confidence.

17 Jun 2026

Today's links look at how to use agents while maintaining agency, why the sign-off layer is becoming the real engineering system, and how tech debt keeps you stuck in pilot purgatory.

16 Jun 2026

Today's links look at whether languages are more than just syntax to devs, why AI demands more engineering discipline now, and why you need to invest in downstream automation.

15 Jun 2026

Today's links look at why "skill fatigue" is real, how to rethink your build versus buy strategy, and how to differentiate a vibe coder from a software engineer.

12 Jun 2026

Today's links look at why it's crunch time for Java modernization, how to teach agents to recover from lost memory, and why the new Colab CLI matters.

11 Jun 2026

Today's links look at why your AI strategy has a trust problem, what loop engineering is all about, and why the PaaS market keeps bundling and unbundling.