Video · 2026-08-19 · 57m · 5 moments

My Coding Skills Are Atrophying

✦ AI generated

timeline · colored by role

Mechanism◆

Not actively writing code causes real cognitive decline as neural pathways become inaccessible, not just a subjective feeling of 'brain drain'.

Scott Tolinski explains that the 'brain drain' from AI coding is real; by not actively participating in the coding process, developers' neural pathways for those skills weaken.

transcript

Scott Tolinski: the brain drain of letting AI do it for you is not just fake. It's not just something that we all feel. It's quite simply by not doing those things in repetition and by not allowing our brain to understand these processes, your brain is actually letting those neural pathways go away or they're there those pathways are uh becoming inaccessible. So it's not just like a feeling of brain drain. By not being actively involved in an active participant in your code or the process by which things are doing things, your brain is in fact not using those skills anymore.

Claim◆

AI latency-induced context switching destroys flow state; the best antidote is limiting concurrent projects and remaining an active participant in the code.

Wes Bos explains that AI coding makes flow state difficult due to latency that triggers distraction, and recommends working on one thing at a time to stay engaged.

transcript

Wes Bos: working with AI to me has become this where I am working on project I'm typing something and that latency in the time at which I get a response back from the AI is long enough that I go to Airbud mode and I look at something else and then I go to this and I go to something else... So, the people that I talk to who do the best job of maintaining a flow state or a non-scattered viewpoint don't work on 13 different things at once. they work on one thing at a time. Uh and like so that is certainly part of the thing that I think needs to be a part of your workflow if you're looking to regain that flow state because it is so easy for somebody whose attention can get pulled here or there to be like well I now that I'm waiting for this thing to finish what I can start something else or go somewhere else or do something else or whatever when like that becomes that turns you into a non-active participant in the code.

provides context · 1

Claim

The AI subscription market is highly volatile, and developers should prioritize portable tooling setups to avoid lock-in as providers and pricing change.

Scott Tolinski recommends building a portable setup using symlinks for skills, agents, and configs to avoid vendor lock-in and adapt as AI models evolve.

transcript

Scott Tolinski: the one thing that you can do, Tim, that is uh beneficial to you is get a setup going where you can make your stuff portable. Because let's say I want to do claude code for a month... you can just sim link everything up all your skills and all your agents files and whatever you you can set up systems to do that... as long as you can make your stuff portable and your processes portable then you can move from one harness to another harness, from one agent to another agent, model, whatever. And to me, that is like the major unlock rather than uh like really investing all of my time into one harness and one model.

Definition

Migration pressure on a tool usually stems from ecosystem decay—hiring difficulties, drying libraries, and integration gaps—rather than the tool becoming technically obsolete.

The hosts discuss when to drop a tool, agreeing that the tipping point usually comes from external factors like hiring and integration support rather than the tool itself failing.

transcript

Wes Bos: My question is where the pressure really comes from. Is it the tool itself getting worse or is it everything around it? Hiring libraries drying up, CI integration, the can someone else pick this up in 2 years problem?... there is a sweet spot in between hanging on to the good old trusty faithful thing that works well... and then it starts to creek, starts to get older, and you start to say, 'I can't hire for it.'

Fact

Web components and modern APIs like View Transitions are the best path to achieving high-quality, app-like UX in traditional server-rendered multi-page applications.

To achieve React-level UX in traditional MPA stacks, the hosts recommend web components and modern browser APIs rather than trying to replicate complex JS frameworks.

transcript

Scott Tolinski: Well, my gosh, Lean, I would say web components. Web components. Hey, web components. Uh, web awesome, I think, is a good uh good thing there to to give a check a try. Part of the reason why traditional multi-page applications don't feel well is because you're missing some of the stuff that you you get by just swapping out one little part of the page or one little tiny island of interaction without having to uh wire up unique JavaScript for each things. And things like uh the page transitions API can now do multi-page applications in some browsers could give you a level of like polish that you're not getting.

provides context · 1

Highlight slides

'Brain Drain' Is Real: Neural Pathways Weaken Without Practice✦ from: Not actively writing code causes real cognitive decline as neural pathways become inaccessible, not just a subjective feeling of 'brain drain'.AI Latency Kills Flow State✦ from: AI latency-induced context switching destroys flow state; the best antidote is limiting concurrent projects and remaining an active participant in the code.The Antidote: Single-Tasking✦ from: AI latency-induced context switching destroys flow state; the best antidote is limiting concurrent projects and remaining an active participant in the code.

Related episodes

Article · Lenny's Newsletter

OpenAI’s Head of Design: This is the best time in history to be a designer | Ian Silber

→

Article · Simon Willison's Weblog

github-code Web Component

→

Audio · CoRecursive

Story: David Heinemeier Hansson, Software Contrarian

→