seer

I think this is indeed the direction, but am wondering is anyone actually trying it right now? Love to explore it.

I think for now the answer to adaerlangsql+++ is just Rust - but its semantics have always been about human usability.

Humans have this thing where we can only keep 7 things at once and have to juggle our context, but recall seems much more broad than model contexts.

If we were to start designing a language and toolkit based around LLM context sizes, limitations and quirks - how would that look?

Also terseness and operators are one thing, but what if it’s the other way around - English words etc - can have a much bigger vocabulary of things that can be done but can easily be generated on the fly.

For a new LLM language realistically you would need to probably either write or convert a shit ton of software to it first if you are to train an adequate model on top of it …

And also what about “IDEs” for agents? So far they are clumsily using human tools - just looking Claude string along bash loops and greps to try to eek out meaning form code _it itself had written a week ago_ is frankly frustrating.

There must be a better way and some rich agent specific tooling for code exploration and verification.

Anthropic stance has always been “yeah we tested them all, grep is better” but I still am not convinced - just by the fact that it constantly fails one or two attempts before “figuring it out” yeah it’s fast anyway but it could be _faster_ and could allow the agent to “think” in code space more and not have to validate itself at every minute step …

Funny how all of this changes with agentic development - once you setup your agents and the initial productivity euphoria , most of the time you spend in deep brainstorming and leave agents to “handle the details”.

Which is both liberating - a lot less punishing mentally and not requiring so much flow focus, but on the other hand you have to keep more loosely held context in your head.

And agents need you attention constantly throughout the day, you leave them working overnight. While you are not in the office etc, and it’s hard to justify when an agent just needs “a quick look” or some confirmation and direction. You leave being paged all the time so you never really relax.

I think a lot of developers will end up feeling what founders actually go through, for better or worse, but also without actually getting the golden ticket option founders usually hold.

It’s just a very different mindset for work, though maybe requiring less chemicals to actually make work

Isn’t “sandbox capitalism” - real capitalism and what the us is trending towards is the “Wild West” capitalism with monopolies, trusts, and all the other anti competitive things that arise in the free market if there is no external force that tries to keep the market “fair”?

I mean this is not the case for either, but the systems are setup to drift in these directions - a market needs some sort of regulation to keep things sane, otherwise you end up with standard oil and the like.

You want the people with guns controlling things to be at least nominally interested in “the people” and produce more than just “return on investment”.

Not saying that China is like this or that US is bad, but saying that China is bad because it is “controlled capitalism” is kinda missing the point of what it should be in the best case.

I always try to think how product owners deal with this. I used to write a lot of code, so did my colleagues, but the product owner was the one driving the effort - they definitely didn’t know how to code, but they “shipped” results regardless - they develop relationships with individual engineers, make sure to understand their strengths and weaknesses, insist on automated testing and then hand verify results, whenever on assumption the engineer expressed but turned out to be not true they push to extract the uncertainties.

Slowly build understanding of what is possible, push back on engineers suggestions and don’t approve stuff blindly, but relying on results to guide them not the implementation.

It is very very similar feeling. -you need to know your models and harnesses (e.g. your … team). You don’t need to know the specifics of the code but doublecheck with different personalities if the assumptions are correct, test them.

And ultimately shipping bugs is a given , human engineers do it _constantly_ it’s the resilience systems you build around that which saves you, tests, canaries, rollbacks, feature flags, everything we already know how to do - just use that and imagine yourself the product owner.

I apologise for the tangent but doesn’t this “future” feel off? We’re improving the experience with navigating all of these noisy, nasty, ugly, toxic and dangerous car sewers (aka car streets) that actively robs us from our humanity and leads to isolation?

This all feels kinda like the Soviet Union building a shit ton of concrete “monuments” everywhere, pretending that was what people “wanted”

The tech is truly impressive but I wish we were able to step out of the local maxima of “optimising this shitty experience until it is less shitty” and start taking “optimising live in cities so it feels nice again” as an actual goal?

Some countries seem to have stumbled on it and the science of urbanism is progressing, but man if that amount of ingenuity that went into self driving cars had went into green and efficient cities…

I thought a second opinion for big life affecting things is just standard practice.

A friend of mine was diagnosed with leukaemia, and the doctor literally told her - get your affairs in order, you got 6months to live, tops.

She spend a day in a daze not knowing what to do, and her parents were like - get a second opinion.

That one turned out negative. The original doctor was like - ah sorry, my bad.

The switch in mindset I experienced was to go from "Working on a feature" to "curating a board" to "overseeing angets curating boards".

So that I'm not just relying on one or a few agents to accomplish a task. I work on the whole product and just drive the direction of tickets and their "what done looks like" requirements, and let the machine organize things on its own, usually 10-20 independent agents (each spawning their own subagents). I think i might have capacity to handle more, just don't have the subscription tokens to fuel it and the budget for the hardware to run it.

Haven't worked in BigCo so no idea how that compares to their tooling. I'm sure it works much more efficiently and on a bigger scale, just being curious about how it all plays out.

Hmm I feel like the social media aspect would be way more useful for democracies in general.

I remember reading about Switzerland’s government service that tracks each politician, their political actions - which positions they take, which laws they promote, etc. and comparing that with which companies and investments they, their famalies and their close connections have.

All quantifiable data. I always thought something like this was _essential_ for any democracy in general - follow the money and “qui bono” and all that, thought it would be cool to build one. But of course prohibitively expensive for individuals small teams. But with all the ai - this could definitely be a thing.

If it all boils down to “verifiable corruption rating” as a single sound bite, it could seriously affect public opinion I feel. Hope someone gets on that idea someday.

To be honest I’m not sure - the tooling around it has processes to start things, and gourds around tool usage to remind Claude of proper procedure.

But ultimately I find it a silver lining more than a problem. I am building and changing tooling so fast, and the models tooling itself is changing so fast, that by the time I establish a workflow, I either A) have solved the original problem so no longer needed, B) have built other tooling that distorts the picture, or C) the labs have released something new and cool that changes the equation drastically so your previous setup is gone anyway.

For example I built some procedure on how to split work between fable and opus and sonnet to optimise token usage, then opus 5.5 came along and made all of this obsolete.

Not per “token usage” mind you, opus is still expensive, but more “effort per task time finish” kind of metric. If I want to close things faster while still being relatively within budget, it is just easier to send all to opus - it will now just solve thigns quicker so you still use similar amount of tokens / money, but quicker, which is a trade I’ll make any day.

And every time “the system is working” I either just stumble onto the next resource constraint, or I feel uncomfortable and start refactoring to make things better and inevitably destabilise the whole “machine”

Hmm how does your custom harness compare to Claude’s?

I’ve tried the various alternatives (opencode, pi.dev) but for “have several physical machines, each running an orchestrator and 5-10 agents, and each agent spawning its own subagents to do work in parallel, all of that on a tanky GitHub project boards” level of work only Claude was doing work dependably.

I can have a few orchestrator leads which I talk to and brainstorm, and let them file tickets, organize them and assemble teams of agents to work of the various problems. And all of that with just a smidge of custom tooling on top of Claude Code’s default.

Plus don’t have a budget to spin all of this on cloud runners so I made sure to do as much as I can with local work on several machines I own. Was quite easy to achieve with Claude.

Do you have something similar built on top of a open source harness, or is it from scratch?

I think the new batch of engineers are going to think faster on their feet, but less precise.

We humans are _amazing_ pattern matching machines and learning “if you do this thing this other thing occurs” is literally what our brains were built for.

It’s just the current few thousand years of technical progress that have been a departure from this reality - we found out about analytical thinking and that by deeply analysing something you can do a shit ton more.

If you look at scientific progress before Industrial Revolution it’s all trial and error between patterns. And I feel like the new crop of devs are silently settling in into this very human and familiar pattern.

On the one hand I think it will probably be good for the “mental health” of people in this profession - you become more like an alchemist acquiring bags of tricks on what works, sharing incantations with others, because sitting down and learning _why_ it works in detail would be a time luxury few would be able to afford.

But that would mean the people under the hood that build the LLMs and who debug the really broken system would be charging the premium.

I honestly don’t think that in and off itself would be such a bad thing.

What I personally am really scared about is the price for labour that is sharply declining. I don’t worry about ASI too much, but when you read history it is absolutely full of stories of great human suffering because “labour was cheap” no need to improve working conditions because there is always someone else willing to take it up.

We had the Industrial Revolution _because_ of the plague making labour expensive and driving the benefit for innovation. Before that nobody really cared. With LLMs doing things becomes vastly cheaper, so the labour itself becomes cheap with it.

Now looking at the state of the engineered world - how absolutely shitty everything is (as opposed to what it technically could be) - there is a shit ton of stuff to be done so the work should be there to fill it. But maybe the capital will move in another direction - nobody can predict the future, I just hope we don’t end up in a dystopian hellscape for the sake of my children.

Honestly Claude code with agent teams and opus 5.5 is good enough for this by default no need for a custom harness / skill

You are right, Indians travel a lot, no other nation does that at all.

And calling someone stupid is right and generally accepted way to lead a fruitful conversation, thank you.

Modi has been the great hero that India wanted and is deserving of every accolade he and his close advisors bring to bear.

I have been now convinced by your argument and steely demeanour. It is truly a joy to talk to someone so charismatic and I feel elevated just from reading these messages.

And you are absolutely correct in assuming that India’s current leadership has no problems, they sample all possible solutions that could be done and select only the best one for their entire people, pulling absolutely everyone from poverty and policing that no entrenched elites capture any of the wealth of the nation.

Of course there are NO protests in this model of a country where the young feel empowered in their future, and even if there were, they would be met with only joyous approval by the elites and all measure would be taken to incorporate any complaints raised by the people to make everybody prosperous.

And elections are absolutely fair and there is never even a hint of any fraud or corruption, because if there ever was that is for sure the cause of evil Chinese spies or British investigative journalist saboteurs.

It is courteous people like you who make me stay in the country and provide my totally undesired technical expertise to bring the living standards of everyone up and pay my incredible fair and not at all inflated taxes for doing so.

I will from now on shut my fuck up and praise the unerring elites at the top.

I remember the old discussion between clojure programmers and someone who works in a strongly typed language. The argument was “types allow you to be certain that if you change some code, an unrelated but affected part of the codebase will also be flagged and it is not missed” the clojure guy response was “why would you ever design a program where a change in one place affects the other”.

I do think this is an unrelated win of functional languages that hasn’t yet been “discovered” by the vibe coder crowd - FP’s whole premise was that it makes your code depend on much much less things so you can “fit it in your head and reason about”… that’s like the perfect sweet spot for agentic as well, we just haven’t seen tools utilize that in earnest.

I wonder what the future of all three is mode bun and deno, their original goal has always been to be able to code in one language across frontend and backend, and typescript itself is a hell of a good language too.

But now that people seldom even look at the code, does it even matter? Our shared language now is English - across tests, frontend, backend, product briefs and design systems.

I have written a crap ton of node code before, because I liked it, but now I do everything in specialized stacks where each tech is chosen to best fit its environment. Backend - use Go, frontend - react/svelte/custom ,game simulation code - C#.

I just debate what would be best fit with an agent, do a few pilots to prove it and just go. Languages I’ve never coded in are super fast to execute - just insist on following best practice, modern conventions and do some spot checks against o(n) problems, architecture and parallelism - and it works, and vibe coded codebase with strict linting rules vastly outperforms fine tuned code in a shared language (node).

I honestly don’t see a bright future for them, and tbh I was surprised at Claude’s migration _to_ bun - why not just make the jump to something like ocamel that would let their agents have even more performance, context and linting tools, but I guess if they did it once the will do it again when they feel like it.