Right now the AI race has a lot of similarities to the dotcom bubble. The subject is packed with risky loans based on huge debts. Those huge debts are expecting to be paid as AI becomes profitable, but AI companies are largely loosing money.
All those loans and infrastructure create the burden of sunk costs leading to a desperate need to succeed.
The people feeling that desperation are the same people who own the largest marketing, news, and social media networks in the world.
As a result, there’s a lot of hype around AI. A lot of “kool-aid,” and everyone wants you to drink it. If you drink the kool-aid, that means you’re also bought into the problem. You also need it to succeed, thus making their problem into your problem.
I explained to him that mature, professional use of AI is going to wind up following a similar path to data engineering. It’ll start with bullshit standards, “prompt engineers” and the like, but eventually SE disciplines are going to define who makes best use of AI. You’re going to have niche use cases for daemon AIs, local LLMs, and remote models. You’ll have stronger frameworks around session management, context management, agent permissions, …
It’s not going to be like this forever, “dump all your shit into our web upload and let the AI figure everything out in one go.” It’s going to become more fragmented, bounded, dare I say deterministic… orchestratable.
Then I told my boss, it would be better if he could frame his excitement around these future use cases… so we can skip the kool-aid stage and get right into the good stuff.
He agreed, until about a week passed. Then it was AI hype again.
The 3rd or 4th “industry expert” tells them that things are “moving fast” and things that were impossible months ago are now reality. It’s designed to make them distrust their own subject-matter experts. They thing, ohh POTV, they’re just not educated and up to speed.
Yeah. Local LLM stuff is great when you want to shove a huge pile of documentation into a model trainer and make a more intelligent search. Two of my vendors have implemented it, and it’s more useful than a traditional indexing search tool, though you do have to verify the results (which is not much more effort since with a search you’d have to skim the document to find the info it matched anyway).
But for general “do everything” tool, yeah no. It can’t read and understand your entire database, codebase, business process, etc.
Honestly, I’ve had a rather interesting experience with AI. I was very adverse to LLM usage at first. Later I sort of figured out that I was more adverse to the energy around AI than I am AI itself.
I knew the models sucked at large tasks. Trying to get an edge on the matter though, I started asking myself, how can I get the model to perform better? I figured I could pass over the AI hate stage and get right into the AI professional stage… at least a head start.
So I began experimenting with local LLMs, LLM harnesses, and various governance tools like jai. I decided against Claude Code and Cortex because they’re provider specific — instead using OpenCode so that I can use whichever model I desire. Then I began building out a SKILL.md repository for tightly scoped tasks like change-review, security-analysis, refactor, architecture-review, grill-me, feature-design, …
I’m still thinking through some of the project needs. I want something that lets an agent work, while treating the agent as a kind of helpful adversary. You should be able to configure workloads that designate models, context, available tooling, skills, permissions, session length, inference level, acceptance criteria, and human-review stages. It would also allow for session switching, model switching, agent deliverable handoff to another agent, … not to mention, your VCS should know and respond appropriately if an agent ever pushes code. Don’t trust it by default.
These workloads should be version controllable, benchmarked, …
Anyway, a lot of that is speculative. Just where I’m at now, controlling context and skills manually, I’m already seeing much better results.
And no, I don’t have the AI do everything. I just find smarter ways to decompose “everything” into much smaller tasks that are easier to review and scrutinize.
But also, I push for local model usage in my organization. I don’t want my success to mean success for the AI companies. Fuck the AI companies.
I was forced to dogfood it. I found that for my specific needs, it made me super productive. I generally make Claude write Ansible jobs, I store all my secrets in a vault that it never gets access to.
It can do tremendous amounts of work at my command in relative safety as long as i provide it protected tools.
Now, that said, I burn a hell of a lot of tokens moving at that speed. When the ass falls out of the market, i’ll still have all the ancible stuff I can reuse.
Neither Claude code neither codex is actually vendor specific, they just don’t tell you that you can config other providers, including local
However opencode is pretty nice too, so if you like it, use that. I personally find that opencode with GLM 5.2 or Kimi K2.7 isn’t actually that great, it’ll hallucinate more than Claude code or Codex with their respective first party models. I think it’s the models themselves rather than opencode itself though, as when I use GPT for planning and hand it off to deepseek flash to do the actual work, it’s more or less fine.
I told my boss this:
I explained to him that mature, professional use of AI is going to wind up following a similar path to data engineering. It’ll start with bullshit standards, “prompt engineers” and the like, but eventually SE disciplines are going to define who makes best use of AI. You’re going to have niche use cases for daemon AIs, local LLMs, and remote models. You’ll have stronger frameworks around session management, context management, agent permissions, …
It’s not going to be like this forever, “dump all your shit into our web upload and let the AI figure everything out in one go.” It’s going to become more fragmented, bounded, dare I say deterministic… orchestratable.
Then I told my boss, it would be better if he could frame his excitement around these future use cases… so we can skip the kool-aid stage and get right into the good stuff.
He agreed, until about a week passed. Then it was AI hype again.
The 3rd or 4th “industry expert” tells them that things are “moving fast” and things that were impossible months ago are now reality. It’s designed to make them distrust their own subject-matter experts. They thing, ohh POTV, they’re just not educated and up to speed.
Yeah. Local LLM stuff is great when you want to shove a huge pile of documentation into a model trainer and make a more intelligent search. Two of my vendors have implemented it, and it’s more useful than a traditional indexing search tool, though you do have to verify the results (which is not much more effort since with a search you’d have to skim the document to find the info it matched anyway).
But for general “do everything” tool, yeah no. It can’t read and understand your entire database, codebase, business process, etc.
Honestly, I’ve had a rather interesting experience with AI. I was very adverse to LLM usage at first. Later I sort of figured out that I was more adverse to the energy around AI than I am AI itself.
I knew the models sucked at large tasks. Trying to get an edge on the matter though, I started asking myself, how can I get the model to perform better? I figured I could pass over the AI hate stage and get right into the AI professional stage… at least a head start.
So I began experimenting with local LLMs, LLM harnesses, and various governance tools like
jai. I decided against Claude Code and Cortex because they’re provider specific — instead using OpenCode so that I can use whichever model I desire. Then I began building out a SKILL.md repository for tightly scoped tasks likechange-review,security-analysis,refactor,architecture-review,grill-me,feature-design, …I’m still thinking through some of the project needs. I want something that lets an agent work, while treating the agent as a kind of helpful adversary. You should be able to configure workloads that designate models, context, available tooling, skills, permissions, session length, inference level, acceptance criteria, and human-review stages. It would also allow for session switching, model switching, agent deliverable handoff to another agent, … not to mention, your VCS should know and respond appropriately if an agent ever pushes code. Don’t trust it by default.
These workloads should be version controllable, benchmarked, …
Anyway, a lot of that is speculative. Just where I’m at now, controlling context and skills manually, I’m already seeing much better results.
And no, I don’t have the AI do everything. I just find smarter ways to decompose “everything” into much smaller tasks that are easier to review and scrutinize.
But also, I push for local model usage in my organization. I don’t want my success to mean success for the AI companies. Fuck the AI companies.
I was forced to dogfood it. I found that for my specific needs, it made me super productive. I generally make Claude write Ansible jobs, I store all my secrets in a vault that it never gets access to.
It can do tremendous amounts of work at my command in relative safety as long as i provide it protected tools.
Now, that said, I burn a hell of a lot of tokens moving at that speed. When the ass falls out of the market, i’ll still have all the ancible stuff I can reuse.
Neither Claude code neither codex is actually vendor specific, they just don’t tell you that you can config other providers, including local
However opencode is pretty nice too, so if you like it, use that. I personally find that opencode with GLM 5.2 or Kimi K2.7 isn’t actually that great, it’ll hallucinate more than Claude code or Codex with their respective first party models. I think it’s the models themselves rather than opencode itself though, as when I use GPT for planning and hand it off to deepseek flash to do the actual work, it’s more or less fine.
I suspect behind the scenes, the first parties are sending your requests to multiple targets and sending you back quorum.