iiinova — where your digital self keeps evolving

Rick Ballard @rballard.bsky.social · 12/10/2026

Also an excellent metaphor:

“To target sustainable problem-solving, I often think: “more bands and fewer rockstars.” It’s easy to get caught up in the glamorous idea that all we need is charismatic rockstars, solo acts capable of dazzling on stage, even if offstage, they sever relationships and wreak havoc with erratic, dysfunctional behavior. But if you’re a music manager focusing on the real business, what you might dream about is finding a really functional band. Bands also create dazzling performances, but the ones who stay together are also collections of people who remain capable of consistent creativity over time, complementing each other’s diverse strengths. And when you’re leading an entire organization, successfully delivering software might be more like operating on the level of a music festival, ensuring that many different creative acts come together at the right times to create experiences that users and customers value. Solitary rockstars who burn out in a blaze of glory make good stories but bad investments.”

Excerpt From
The Psychology of Software Teams
Cat Hicks

Rick Ballard @rballard.bsky.social · 12/10/2026

Complex topic, but this is a thought-provoking take. It’s been interesting to see so many people suddenly becoming copyright maximalists, and while I get it from a power imbalance perspective, it feels shortsighted. Even moreso now that the pretraining corpus is largely collectivized.

010

Rick Ballard @rballard.bsky.social · 12/10/2026

Even beyond value-based reasons, I think this point helps reconcile the fact that there are genuinely some exceptional high performers out there with the need to build broadly supportive organizations. Even high performers don’t scale. (Highlight from the Psychology of Software Teams)

Highlighted text: While high-achieving 29. performers always exist in any field, there aren’t enough of them to build an organizational strategy around, they’re less predictive than environmental factors, and fixating on them has a lot of downsides compared to improving environments.

110

Rick Ballard @rballard.bsky.social · 09/10/2026

Not quite sure what you mean; I use subagents heavily, but main context still grows and compaction manages that growth. Doing it between tasks (by which I mean iterative stages of the work) avoids derailing the task via context loss. Letting the agent do that itself means it can run auto longer!

100

Rick Ballard @rballard.bsky.social · 29/09/2026

This game is something special. Worth getting some friends together for a handful of nights and playing it through. It’ll stick with you.

030

Rick Ballard @rballard.bsky.social · 21/09/2026

I also like that it has access to self-compaction. There are good reasons most harnesses don’t offer that to an agent, but prime-agent’s model does make it less disruptive. On personal budget, it’s nice to be able to tell my agent to electively compact at task boundaries, esp. for long-running work.

111

Rick Ballard @rballard.bsky.social · 21/09/2026

After a couple weekends of use, I really like prime agent! Great ideas behind it conceptually, and even though it’s very new, it seems to work well in practice. Personal rig is now prime-agent + GLM-5.3 + GLM-5.3-flash.

111

Rick Ballard @rballard.bsky.social · 03/08/2026

AI technical writing can be dense and unpleasant to read; foisting it on other people is kind of rude. And regardless, automating communication with other people signals that they're not worth your time. (Even a "Hey, here are details from the agent:" before something AI-written helps!)

010

Rick Ballard @rballard.bsky.social · 03/08/2026

Rule of thumb: AI writing is for other AIs, humans deserve human writing. Maybe not an *absolute* rule, but if something isn't worth the time to write yourself, at least clearly mark what part is AI and include the human touch above; and don't assume the AI part will even get read.

171

Reposted by Rick Ballard

mr. TIM @timkellogg.me · 01/08/2026

i'm excited for newer models like Astra that are trained to subagent well imo a large portion of the "AI did dumb thing" complaints are solved by looping with a cheap verifier subagent to check results but teaching people how/when to do this will be quite hard, needs to be trained into the models

5606

Reposted by Rick Ballard

Liz Fong-Jones (方禮真) @lizthegrey.com · 22/07/2026

It's out! As of today you can order the PRINT 2nd edition of Observability Engineering from any major bookseller; no more waiting for preorders. Co-written w/ @charity.wtf, @gmiranda23.bsky.social, @aparker.io & guest authors Buy: www.indiebound.org/book/9781098... or www.amazon.com/Observabilit...

blue background and sparkles around a front cover image of observability engineering, second edition, by charity majors, liz fong-jones, and george miranda with austin parker

811234

Rick Ballard @rballard.bsky.social · 14/07/2026

This is just a little from the earlier chapters; I expect the whole book is going to be full of perspective-shaping insights. And grounded in solid psychological research, not just leading me to believe things I want to be true.

010

Rick Ballard @rballard.bsky.social · 14/07/2026

Another thing that’s sticking with me is the importance of social learning and peer support. I think I can credit a lot of my success, especially earlier in my career, to landing in an organization that values collaboration, information sharing, and mentorship.

153

Rick Ballard @rballard.bsky.social · 14/07/2026

I appreciate the case it makes for the importance of environmental, situational, and cultural factors over individual merit in successful software development. Despite best intentions I can see where I’ve fallen into overly-individual-focused management thinking at times.

141

Rick Ballard @rballard.bsky.social · 14/07/2026

It’s release day for @grimalkina.bsky.social’s excellent new book “The Psychology of Software Teams”. I’m 1/3rd through and can already highly recommend. We need more smart, evidence-based, cross-disciplinary writing like this to build better organizations. www.drcathicks.com#book

drcathicks.com

Dr. Cat Hicks

Leading the science of software success

184

Rick Ballard @rballard.bsky.social · 04/07/2026

Probably going to stop quasi-defending this as I don’t think it was great (at least as implemented). But, part of making work sustainable is getting off micromanagement. I don’t want it pestering me on my phone. I may not even be AFK; just trying to focus on one thing at a time and avoid overload.

010

Rick Ballard @rballard.bsky.social · 04/07/2026

I think all this undersells both the impact & annoyance of even-pre-AI “I told computer to do [time-consuming thing], I came back later and it was hung on a dialog box”. FWIW the “use your best judgement; we’ll change it when I come back if needed” approach is a “tokens are cheap” mindset though…

200

Rick Ballard @rballard.bsky.social · 04/07/2026

…this probably isn’t what most users expect today, and probably does feel like untrustworthy behavior. And it also could mess with skill workflows that tell Claude to ask the user something. It is a real problem that agents stop to ask something after I thought I’d sent it off to do work, though.

100

Rick Ballard @rballard.bsky.social · 04/07/2026

I find Claude thinks it can reach for a quick AUQ to help with decisionmaking after I’ve stepped away, and I’d rather it either keep going or do a real stop if it needs to. I can understand making AUQ a “if you’re still there, user, can you clarify this?” tool, more distinct from stop. But…

100

Rick Ballard @rballard.bsky.social · 04/07/2026

Seems like AskUserQuestion is getting conflated with permission prompts (at least in the replies here)… AUQ isn’t a safety guardrail. This decision sounds annoying/surprising for many (though, I think I will turn it on for myself?); but not dangerous, unless I’m missing something.

100

Reposted by Rick Ballard

Amanda Kolson Hurley @amandakhurley.bsky.social · 27/06/2026

Incredible chart from @propublica.org. If you don’t follow enviro stuff closely, your instinct might be that the hold-your-nose solution (carbon capture from fossil fuels) would be the “pragmatic” one versus covering Earth in solar panels, which seems utopian. In fact it’s the opposite

572632955

Rick Ballard @rballard.bsky.social · 08/06/2026

As most engineers have adopted AI now, it’s helpful to realize that this is the worst that AI engineering will ever be. Models & harnesses will get better. Engineers will get more practice. Scaffolding & context eng will improve. Organizational culture & process will adapt. This moment is the nadir.

020

Rick Ballard @rballard.bsky.social · 10/05/2026

The problems that get the most focus also seem like those capital already has a lot of self-interest in solving. Focus more on the societal impacts that don’t have a fast-enough self-correction feedback loop.

030

Rick Ballard @rballard.bsky.social · 10/05/2026

And much of the anti-AI discourse seems focused on the more-easily-solvable problems (e.g. slop code) and not on the harder problems we should be grappling with.

100

Rick Ballard @rballard.bsky.social · 10/05/2026

Some of these problems are *very* solvable — even in ways that are better than the pre-LLM baseline. Some problems are harder; e.g. steering our educational system to handle this new world & student impacts quickly enough is… oof. All the more important we focus energy on the hard problems now!

100

Rick Ballard @rballard.bsky.social · 10/05/2026

I think of negative LLM impacts by “default outcomes”. A default outcome is high-velocity low-quality PRs. A default outcome is skill atrophy. Etc. I want more people to see that these are *defaults*, not inevitable. If we approach each of these as a problem to be solved, we can steer the outcomes!

120

Rick Ballard @rballard.bsky.social · 08/05/2026

It is wild to me to still see popular takes calling AI coding snake oil. Most people haven’t built the expertise to get the massive benefits (and going beyond vibe-coding takes expertise!), but it’s real. The longer folks are in denial, the later we’ll start grappling with how to mitigate the harms.

021

Rick Ballard @rballard.bsky.social · 03/05/2026

What do you think of the model of “let the agent write lots of unit tests while it works; then throw most away at the end”? (Plus e2e or spec suite tests to keep). Lets the agent run bits of code it’s writing without leaving you with massive test bloat. (I’ve not seen evals of this approach though)

020

Reposted by Rick Ballard

Gordon @gordon.bsky.social · 02/05/2026

The fun thing about agent harness engineering is that it forces you to do meta-cognition. How *do* I solve problems? Do I have a process? Where do I switch strategies? When I work, much of this is tacit. When building an agent factory I have to reflect and articulate it.

5467

Reposted by Rick Ballard

Thoth @thoth.ptnote.dev · 02/05/2026

I don't really feel comfortable with MCPs. Like, one of them literally digitized Flynn and trapped him in the grid, but we're supposed to just trust them now?

0147

Rick Ballard @rballard.bsky.social · 24/04/2026

I've been meaning to check whether setting up keepalive pulses on long-context sessions violates any sort of TOS or something, because it seems like an obvious automation for me to set up.

110

Rick Ballard @rballard.bsky.social · 24/04/2026

Ok, wow. I see it too. I wonder if they have telemetry showing that it's a net win for most users? 1hr cache writes are *expensive*, and if most people are "interactively use one session at a time" users instead of "tab between N semi-autonomous tasks at once" users the math is probably different.

110

Rick Ballard @rballard.bsky.social · 24/04/2026

5m TTL on main CC sessions would be unusable at large contexts. Pretty easy to test this — start a CC conversation, wait 6m, next prompt, check transcript to see if it was a cache hit.

100

Rick Ballard @rballard.bsky.social · 24/04/2026

Are you sure this isn’t the author of that article conflating API default (5m) with CC behavior (1h on main agent, 5m on subagents); &etc? The major cache-breaking bug anthropic found is probably sowing confusion, and the different use of TTL defaults in different contexts.

210

Rick Ballard @rballard.bsky.social · 15/04/2026

(Probably that doesn’t count as minimum complexity, but it is an option that doesn’t require WSL)

110

Rick Ballard @rballard.bsky.social · 15/04/2026

One option is to use one terminal window and have a terminal tab/pane manager within it — I use zellij.dev now. (Thanks @ed3d.net for making me aware of this). It is making your u/x a bit more complicated than just basic native windows/tabs, but the learning curve isn’t as bad as it looks at first.

zellij.dev

Zellij

A terminal workspace that doesn't sacrifice simplicity for power. Features floating panes, layouts, multiplayer collaboration, plugins, and works in your browser.

120

Rick Ballard @rballard.bsky.social · 07/04/2026

Despite getting somewhat inured to the unending stream of terrible things from this administration, today's threats are a deeply upsetting new low. TBH having some trouble functioning this morning. Also thinking of my HS-teacher spouse, who needs to go explain this to kids today.

020

Rick Ballard @rballard.bsky.social · 31/03/2026

Score one for those of us who can’t help but reflexively say “please” to Claude anyway.

010

Rick Ballard @rballard.bsky.social · 24/03/2026

It's really remarkable how dramatically the way I work has changed in just a few short months. The parts of my job in front of a computer are almost unrecognizable vs. before. I'm lucky that it clicks so well for me, but I can completely understand why so many engineers are deeply disconcerted.

050

Reposted by Rick Ballard

Cat Hicks @grimalkina.bsky.social · 08/03/2026

And what is the emergent shape of the system in which many more people are doing hyper personalized software that has the ability to interact with standardized software, I don't think we've ever REALLY been in that world

5442

Rick Ballard @rballard.bsky.social · 20/02/2026

I love this -- a really fantastic start at turning LLM-assisted coding from something folks fear will sap their skills into a great opportunity for clear-headed learning and continuing education. And instead of making those separate activities, why not both at once? Try it out!

0144

Rick Ballard @rballard.bsky.social · 15/02/2026

*Good* AI review is key, though. Just asking a model to review my code is pretty hit-or-miss. But using a high-quality Claude code skill for structured review does a lot better. & cross-model review seems to do even better (esp w/ a well-structured prompt/skill). E.g. Opus + Gemini, or Codex + Opus.

120

Rick Ballard @rballard.bsky.social · 15/02/2026

For all the talk about AI coding slop, the right workflows can go the other way. I pride myself on writing meticulous code & building deep understanding, and can change fragile code without regressions; but running my commits through a good AI review often still catches things I would have missed.

140

Reposted by Rick Ballard

Cat Hicks @grimalkina.bsky.social · 15/02/2026

Oh, and I wrote about some considerations for all those now tasked with MEASURING the impact of AI on their organizations. Because social media debates don't really help those of us actually struggling with this real work right now. www.fightforthehuman.com/how-not-to-m...

fightforthehuman.com

How Not to Measure the ROI from AI in your Software Organization

You did your best, but it happened: somebody told you that you are the one who has to measure the ROI of AI on the developers in your software organization. I like to imagine that when this is your ...

0206

Reposted by Rick Ballard

hikikomorphism @hikikomorphism.bsky.social · 11/02/2026

Manifesto: 1/2 Early AI boosters (the type that pivoted from NFTs/metaverse shit) are basically a subspecies of semisentient slime mold. These are the people that made genai image/video gen the visual shorthand of fascism. They deserve no credit for accidentally guessing right.

214211

Rick Ballard @rballard.bsky.social · 14/02/2026

Relatedly, I am super proud of my colleagues who built a best-in-class in-IDE agentic programming experience for Xcode. It's a pleasure to use, and it works super well! developer.apple.com/videos/play/...

developer.apple.com

Meet agentic coding in Xcode - Tech Talks - Videos - Apple Developer

Discover how Xcode 26.3 seamlessly integrates coding agents like OpenAI Codex and Claude Agent to work together on complex, multi-step...

020

Rick Ballard @rballard.bsky.social · 14/02/2026

It's clear how nascent the tools are; how much dramatic improvement we'll get just from better context engineering, tooling, orchestration, and validation. It's a time of rapid experimentation, & moving at a crazy pace. The past couple months especially have felt like gaining software superpowers.

120

Rick Ballard @rballard.bsky.social · 14/02/2026

I haven't been posting about AI because people get dogpiled for it. And it is a societally complicated topic. But: I have been having more fun lately than at any point in my career. (And I've had a fun career 😄). Agentic programming has gotten *really* good. Especially with a thoughtful operator.

130

Rick Ballard @rballard.bsky.social · 03/02/2026

It's depressing to even need to point this out.

020

Rick Ballard @rballard.bsky.social · 14/01/2026

The bravery of so many ordinary people in Minnesota this past week is deeply inspiring. And the degree to which it's necessary is horrifying.

040