CharlieDigital

If a team wants the benefits of velocity with AI, the consequence is volume.

To deal with the volume, the system needs more checks than humans can handle. Additionally, the humans no longer understand the details at the low level of implementation at scale. Therefore, the system only needs to assert that it is following agreed patterns.

So the initial scaffolding needs to be agent friendly. The processes need to be aligned to agents. The tooling needs to be agent operable and put on rails. The patterns need to be documented to be enforced.

Otherwise, teams get volume that cannot be handled by humans. So full speed agentic engineering requires teams to be all-in.

A piece of this is to leverage the harness to normalize or standardize output. Standard patterns, standard architecture, standard UI libraries. These are all much easier to enforce with agents than with humans, but requires extensive platform and harness engineering.

Teams do not really understand this yet. Engineering teams still think their job is to build product faster. By the 10th time you tell an agent "implement issue #15639", you realize anyone could have done that. Engineering's job now is to own the enterprise harness (context layer, ephemeral agent env, ingress channels, auth proxies, MCP servers, quality gates, agent ops, etc) so anyone can build the product. That's what "all-in" means.

Once this system is set up and well executed, it really doesn't matter who writes the initial prompt. An enterprise harness designed to enforce safety, security, and best practices applies to a senior dev or a junior dev or a product owner all the same. All-in on engineering the enterprise harness, not the product.

That's how I see it.

You have to go all in on agentic engineering. And if you're all in on agentic engineering and doing it well, then anyone should be able to build safely and no one needs to look at code. The code should look highly consistent and agents should be enforcing best practices, approved architecture, running security scanning, etc.

The gains at the middleground compromise are not really significant and, IMO, creates frustration and friction.

Works best to go all in. GS should let everyone have CC and GH and focus on making their enterprise harness safer for everyone.

    > A trading floor is a bad place for unreviewed code written by people who can't read it, and the information barrier problem would be reason enough on its own. Requiring that agent-written code come from people trained to review it is a sensible rule for a regulated business.

Jokes on them: devs are also losing the capacity to read the code. Can't mix move fast, high volume with "read code".

Agentic engineering needs to be "all in".

Paseo has been pretty amazing at getting more autonomy. It's the way the agent profiles work. It is effectively a fully adaptive "auto-harness" with just the tiniest amount of config.

Fully switched from TUI 4 weeks back while testing it out for another use case.

Wrote about my setup here: https://chrlschn.dev/blog/2026/10/larping-with-agents-and-wh...

Give and agent the llms.txt and ask it to read.

I have a dedicated agent profile that specifically owns this task and I tell the other agents to only send screenshots to it and ask for feedback.

I think this might be more like a TSMC situation where the rest of the world is always behind.

China will have a critical mass of research, scientists, manufacturing, engineers, and the right government buy in to stay ahead because there's so much demand.

    > It's a shame __________ didn't lean harder into EVs 

That's the story with just about every "legacy" manufacturer because they don't understand the existential nature of the shift; it doesn't feel real while their ICE are still their money maker.

Local models: if they have the hardware to run it.

Open models: we're already seeing this at some scale. https://news.ycombinator.com/item?id=50350004. I do expect to see more of this.

    > ...a problem I've had with RAG - when a fact isn't local to a single patch of text

Is this a problem, though? Vector similarity should surface the relevant, related patches.

    > GitHub ToS actually allows that since April

Do you have more background on this? My team actually ran into some issues with GH when we tried this earlier this year (felt like it was around that same time).

I totally get using AI to make the decks; it's drudgery.

But now that one no longer needs to actually perform the mechanical deck building, how about spending that gained efficiency on ensuring quality of the output? Play it back at least once. Use that time savings to prep to deliver a killer talk.

    > ...my management seems only interested in volume of bits

It's as bad as teams that are measuring by LoC or PR volume.

Whenever I hear this come up, my first question is "how are you matching that to revenue?" and you can see gears churning...

I don't think it's fair at all to blame AI for that; a human made the final call to drop those bombs.

A lot less than you probably think.

My line of work: I deal with enterprise SaaS companies and I can see first hand that engineers are very slow moving with AI, very cautious, and still believe they need to review small PRs.

AI lowers the bar significantly. Just as with kids and essays; it's not the plagiarism or cheating on an essay is new, it's that AI makes it trivially easy to engage in low-value behavior and thus higher incidence.

In this team? Pre-AI, maybe 10% of the slides would be questionable. Post-AI? 90% of the slides were just "meh", broken animations (they never previewed it), and "uh, let me read that slide".

I can understand both the automobile and electricity because those were legit dangerous and worthy of some caution.

    > The Valley doesn’t understand how fed up normal people are with AI. This is the first time ever a technological advance is so cordially disliked, and the Valley is not equipped to detect that. It has no experience with being disliked by normal people.

An anecdote from a recent occurrence. We show up for a team offsite. Several leaders presented decks. Of course, AI generated decks. Some of the animation was broken. Many of the read throughs felt like they were the first time folks saw the decks.

Cannot state how off-putting this was as someone participating in this team activity. The talks I appreciated the most were the ones where people didn't have slides at all and talked off the cuff.

It's not just US manufacturers. Toyota, Nissan, Honda, Mazda, etc. are also falling behind on the technology, expertise, and platform development.

Koreans look competitive. It's interesting that the later the entrant to the market, the faster they are to adapt.

It isn't just US manufacturers.

I'd say Koreans are probably the closest to the Chinese, but Japanese manufacturers have fallen behind. Toyota spent too much time with hydrogen. Nissan? Honda? Mazda? None of them have a solid lineup of electric vehicles.

I appreciate Toyota's hybrids, but that is not the future and every model year that they try to delay, they are missing out on developing the technology, experience, internal expertise, and platform refinement.

Not only that, Musk straight up decided to support Republican candidates who are vocally anti-environmental, anti-electrification, anti-incentives to shift consumer and industry change. He supported candidates and ideology that do not align with the buyers most likely to associate themselves with EVs.

I am not making a judgement here at all and I'm making zero comments on Musk's ideology or character, only that he made a poor branding decision by supporting a voting base and politicians that to not represent the folks most likely to buy a Tesla.

You don't need to argue the intent of the salute; you can just see which candidates he supported and donated to.

(It's fine to debate policy on things like incentives, but that is at least part of the reason why we see China making big leaps here because the government has aligned the incentives with buyers and manufacturers to make electrification the default)

Unions and environmentalism are two orthogonal issues. I'm not sure how you tie those two together.

There isn't even a need to discuss the intent of the salute; just look at who he donated to and who he supported as candidates. The problem is that buyers who want EVs are much, much more likely to align with the opposite ideology.

Yeah, but those ppl want trucks, not an EV "truck"

Worst thing Musk did in context of Tesla is to alienate core demographics for EVs with his politics.

Yes, 100%. There's some recent research that backs this up.

Adaptive Auto-Harness: Sustained Self-Improvement for Agentic System Deployment on Open-Ended Task Stream (https://arxiv.org/pdf/2606.01770) talks about this and the efficacy of role- and objective-based agents on unbounded streams of heterogeneous tasks.

I recently wrote down some of my own experience with this: LARPing with Agents (and Why It Works) (https://chrlschn.dev/blog/2026/10/larping-with-agents-and-wh...) after someone called out my role-based agents as "larping". Short of it: role-based, fully "JIT" agents have increased autonomy in my coding workflows end-to-end. I get very good outputs with few skills required and less human instruction input.

That feels like it's more a function of the lack of a query UI whereas Pg and Mongo have a lot of options.

I used Firefoo for a while and Google should just buy that product and ship it with Firebase. Otherwise, I have no issues with local Firestore emulation; everything works as I expected it with a few edge cases.

Many F500 are already using Chinese models or models derived from them through American providers. Even Microsoft's Decision-1 Jev-like:

https://commandline.microsoft.com/microsoft-decision-1-model...

    > To build Microsoft-Decision-1, we post trained Qwen3.5-9B for fast, single-pass decision scoring and will soon rebase it on other models, including Microsoft AI (MAI) and OpenAI.

Cursor's models most famously are built on top of Kimi:

https://techcrunch.com/2026/03/22/cursor-admits-its-new-codi...

    > Cursor’s vice president of developer education Lee Robinson soon acknowledged, “Yep, Composer 2 started from an open-source base!” But he said, “Only ~1/4 of the compute spent on the final model came from the base, the rest is from our training.” As a result, he said Composer 2’s performance on various benchmarks is “very different” from Kimi’s.

So yeah, Chinese open weight models are already in F500.

I used Ollama with Qwen 3.8 27b on my MBP for a flight to try some experimentation and found that it was shockingly good at one of the more complex PDF processing tasks.

Like seat-of-the-pants good as Opus 5.5 but a bit slower and required just a bit more prompting to get what I wanted (which is fine for an app pipeline).

I have had some success with this by using a distinct agent profile for this role in my agent team.

The agent is instructed to use Laws of UX[0] when reviewing application flows and cite specific laws with changes to the builder agents. Look-and-feel in general still requires some tuning, but I have found that: 1) using a discrete agent in this role that only looks at the flow and is given the end-user context, 2) pointed to a set of guidance has really improved autonomy for UX.

[0] https://lawsofux.com/, https://lawsofux.com/llms.txt

    > however much you distrust the media, it's not enough...

The YT video is not from a Tesla hater, but shows real, structural failures with the CT and the engineering (using cast aluminum instead of steel).

I don't trust any singular source, but looking at evidence in aggregate across many sources, there are clear problems with the CT from material selection to parts engineering to overall capability vs what was sold.

Most cars are and have been roughly the same shape for decades now. Rare innovations here and there, but largely still hold the same basic shapes across manufacturers and markets.

Read the recall link:

    > If the cracks persist and worsen, it may result in the wheel detaching entirely.
    > 
    > ...engineers determined that continued use under strain could eventually lead to separation of the wheel stud from the hub.

Watch the YT video.

Telsa made bad engineering choices on this vehicle. There's no point defending it.

Is it because your world view fractures with a fallible Musk?