- Spend most time prioritizing/discussing what to do.
- Once that's agreed, use Fable 5 High + 5.6 Sol XHigh come up with a design + plan. Agree on the high level plan. (Usually this just comes down to choosing where the change belongs on the spectrum between minimal patch <-> full redesign)
- Use Opus 5 or Sol Med to execute
- Auto-fix bugs and CI until green + thermonuclear review skill x3.
- Manual interrogation of change/nits
- Come up with QA plan and have Codex Computer Use execute on it
- Manually spot check the final result (usually a sizable diff, thousands of lines, complete feature E2E, etc)
I probably spend like $80 a day at least but I produce the output of 3 or 4 2022 engineers and probably at better quality. So it's easily worth it. Would I save money by switching to GLM 5.2 and such...perhaps? IDK. At our scale it's not worth the time spent building the eval harness to actually understand the performance tradeoff.
When I personally want tight code, I have to spend a considerable amount of time adjusting it manually:
- It needs to be trimmed down. In my experience, at least one agent I use struggles to produce minimalist designs, and it's very frustrating
- I need to consider whether there are solutions based on higher-level assumptions, that AIs typically miss
- I need to check whether there are off-the-shelf solutions - AIs like to reinvent the wheel
IMO, software production has become a mass-produced commodity in every sense - it's much more expensive to produce software manually, but the quality is not the same.
> software production has become a mass-produced commodity in every sense - it's much more expensive to produce software manually, but the quality is not the same.
Agreed.
Homer reunites with his long lost brother, who runs car company Powell Motors. Homer is ultimately tasked by his brother with helping to design a car for the “average man” that ultimately bankrupts the company for being wildly overengineered and costing too much ($82,000 in 1991-money).
Here’s the car: https://simpsons.fandom.com/wiki/The_Homer
https://amphetamem.es/meme?id=the-simpsons_02_15_490&text=Th...
[1]: https://pluralistic.net/2024/04/24/naming-names/#prabhakar-r...
Same with a junior dev. They don't write long form spaghetti because they're trying to write more LOC. They do it because not doing it is hard, literally above their pay grade.
I use LLM every day, but they're still completely awful at architecture. I don't think this clear lack of ability is some conspiracy.
I was only able to do that after I had solved multiple related problems in different places and started introducing subtle bugs by accident / had difficulty detecting all edge cases
I've noticed whenever I use LLMs they introduce the same kind of thing but at much smaller scales than I would. They often suggest solving the wrong problem when I prompt them to diagnose specific bugs too. Usually opting for a shortcut that introduces its own issues and ironically calling the proper direction "too complex" when it's really not.
You will really have to weigh the cost of making the software against the expected revenue.
In such markets, what you produce is either worth nothing or worth millions of dollars. For as long as it's the case that well-constructed code (with or without LLM help) is more likely to be in the latter category, the economics of software don't really change.
Even before LLMs, you could've commissioned a half-assed clone of any app you wanted from a 3rd world consultancy for a few thousand dollars. LLMs are basically Bangalore-as-API.
I remember hearing a story that in the past movies were so technically difficult to make that any movie that got made had a good chance to be a profitable hit. But as movies got cheaper to make, more movies got made. Nowadays movie studio execs have to really calculate out the audience and expected revenue for any new movie and balance that against the budget and the cost of the studio's failed movies.
I think a similar dynamic may happen in software
That might be part of it, but I think it also has to do with the reality of replicating and scaling. Hardware or physical goods simply don’t scale like digital goods. There can be hundreds of knock-off physical products that have lower quality and lower cost but serve 90% of the same purpose, because physical capacity for raw materials, construction, labor, shipping, etc. have scaling limits in each market and economy. Digital goods are just so much easier to replicate and scale, so it often doesn’t make sense to buy software at lower quality and lower price if it doesn’t do most of the job. There are still limits of course, and different from physical goods, but I think this is a key reason why software is seen as winner-take-all.
As the LLM's keep getting better at writing software actual coding skills required will decrease which will result in a huge influx of software crashing prices. You can see this in already in open source projects in the last 6 months to a year Jellyfin had a huge influx of custom clients made by different people for fixing their own itches about the official software.
I.e. because software was hard to make and complex to copy you would tend to have "natural monopolies" that were hard to compete with. Who wants to try to build a new desktop OS to compete with Windows? Or a web browser from scratch? Or a new search engine? Etc.
Those and other pieces of software were complex and hard to make. The cost to copy and compete was very high. So one winner took most because that winner was the company who could figure that software out.
But as we can see with Kimi Work and other such things, software is now much easier to copy. Let's say it took $1 billion to make a copycat piece of software with people but now takes $100 million or $10 million with AI. Suddenly a copy and compete tactic makes much more sense than before
For example, with AI it might make financial sense to build a Chinese Native OS instead of Windows. Similarly for Russia, Iran, the EU, and a whole bunch of other places. All of a sudden, Windows might not be the winner take most OS, we might have lots of Operating Systems, with smaller markets and lower profits, which require much more careful financial analysis to stay profitable.
This would be just like Movies, TV Shows, Books or Music. When something works, people relentlessly copy it and different regions put their own spin on the idea. After Iron Man succeeded we had so many super hero movies. Etc. So there is not really a winner take most dynamic in these other digital products. Software may be moving that way
This could not be further from the truth. It might be true if the software works without supervisors in a B2C scenario where everything is self explanatory and nothing can go wrong, movie streaming services for example, and even these have tons of variants and multiple big players racking up billions in market share. So even there "winner-takes-all" does not apply. It gets even worse in B2B software, as there is SO MUCH software thats commonly used and needs tons of supervising by experienced devs. If we had a "winner-takes-all" market a lot more software would already be dead. And a lot of companies would have way less options when adopting new software for something.
Fairly sure software dev was always an iceberg. Most software and most software devs aren't working on horizontal software, but on vertical software, in cost centers. Sales for that kind of software don't scale as much.
When we were making brick houses we needed a few brick factories with thousands of people for million of bricklayers.
For who?
The public? The public has never liked buying software at any price.
Businesses? Businesses need higher quality software when it's relevant to their core competencies, so they hire people instead. Buying competing SaaS or depending too much on AI is throwing the baby out with the bathwater.
Also just remember - minimalist code looks and feels great but customers do not read your code. I have caught myself many times providing "corrections" to abstractions that were already ~fine, just not perfect. The average SWE costs $200/hr. Careful you don't burn $50 worrying about code that will likely be rewritten or can be better abstracted when that's actually needed.
This is a pointless quibble but the hourly rate claim is not true--it's like ~$60 in the USA [0]. Maybe you meant at a specific Org but this is important context when comparing "pricing" between human and AI.
[0] https://www.salaryexpert.com/salary/job/software-developer/u...
Even for a junior making $100K, I have a hard time believe their time is worth less than $75/hr or so.
Edit: Fine, "Senior" is not "Average". But naive salary is not the true numerator.
I have a hard time believe their time is worth less than $75/hr or so.
Pay hits a ceiling, and that ceiling is moving lower regardless of experience. That has nothing to do with AI, but what the market will bear. Hiring counts of humans must increase no matter what. Moving some of the spend to AI reduces the risk of hiring less qualified employees they might have rejected a decade ago.
Wages at the top end are stagnating to subsidize this. That's undeniable.
And this is how I find out I'm woefully underpaid.
We're entering a market of lemons in software. It's going to be interesting indeed.
"Will I benefit from this code being minimalist before [date]", where [date] is whenever you think the agent will be good enough to come back and make the corrections you would make today.
Even as a self-contained unit, you can't step in the same river twice, and on [date] some important details may have seriously faded, both in terms of text that can be mined and also in terms of human "why did we do that" and "what was the reason we did it this way and not that way" etc.
How does a “small startup” end up with a multi million line “legacy” codebase? Something not mathing
Easy! The output of 6 months ago Opus! Which seemed so wonderful at the time.
My observation has been:
- Initial greenfield work by an LLM is fast and very effective with minimal or no human oversight.
- Subsequent work ends up being over engineered and very verbose. Assumptions are made that aren't suited to the problem at hand (for example I find Fable is extremely regex happy where structured data would work much better from a readability perspective.)
- Once code bloats beyond a certain point due to unguided LLM usage, complexity is high enough that only LLMs can operate on the codebase with any economical amount of time.
- Rinse repeat and your code ends up unclear about any state that's not explicitly being tested and verified in QA loops
For some of our products this has been fine, for others it's been problematic. An understanding of your size and reliability requirements will help make the conversation more productive.
Why aren't you guiding your LLM usage? Is that what I said - to spam it and not guide anything? Or to have a careful workflow where you agree on design and maximize your human judgement/leverage?
> any state that's not explicitly being tested and verified in QA loops
As opposed to before, when engineers perfectly reasoned about code behavior from first principals and QA was unnecessary?
You didn't say anything positively or negatively regarding this so I made an assumption that you were using the LLM relatively unguided (e.g. a bit of oversight, not the kind of thing that heavy code reviews used to involve pre-agents.) Feel free to add clarity on your actual usage loop.
> As opposed to before, when engineers perfectly reasoned about code behavior from first principals and QA was unnecessary?
In my experience, most engineers are quite good at reasoning about code behavior for non-QAed code paths. Obviously things fall through the cracks. But I've been in the ground floor of plenty of Big Techs in their early stages before agents and, yes, a lot of initial development had spotty test coverage and yet most of the engineers had good mental models of what was happening. It used to be a very valuable skill to wrap your head around a torrid piece of code with few or no tests but was nonetheless a core piece of your application. Conversely, agentic development can bring cognitive debt [1].
===
This isn't a fight. We aren't sparring over what's right and wrong. I'm just curious how other people use agents in their work as someone who is also now in a startup that uses LLM agents heavily and has no limitations on spend.
I feel like this statement betrays your lack of advanced experience coding with LLMs.
OP's elaboration of the steps they are going through (planning, agreeing on plan, getting one LLM to draft execution plan, approving it, then executing with a separate LLM, then reviewing/testing) made it super obvious to me that they are guiding their LLMs quite considerably as part of their work.
Anyone making blanket statements about LLMs producing garbage is just telling on themselves about not having proper SDLC practices in place.
I find for my own work that I need to read the diff the LLM produces then offer feedback on the diff in its own loop before I am satisfied, and this is after all the unattended QA steps through Codex Computer or Claude MCPs happen. Then auto reviewers come in and then reviewers come in. Of course, at our stage, we rarely have this luxury and it's only reserved for the very core of our codebase.
This is still much less guidance than we used to do for code before agents became popular. Even at Series A companies, before agents, we used to socialize tech specs, get buy-in from multiple engineers, create test plans, etc etc.
> Anyone making blanket statements about LLMs producing garbage is just telling on themselves about not having proper SDLC practices in place.
> I feel like this statement betrays your lack of advanced experience coding with LLMs.
Are we in school debate club? I don't know what's going on lol, I'm just curious how people are using LLMs! Is it just that irresistable to take a cheap shot at each other?
Not that I know of but that's the conclusion I drew from your statement.
It's not a cheap shot unless you took it personally?
I suppose I could have said "the fact that OP's explanation of how they work did not lead you to conclude they were in fact guiding their LLM usage quite a bit tells me that perhaps you have not been working with LLMs in any advanced capacity".
For the SDLC comment I admit it was a broader statement (based on observing people generalizing that "LLMs produce bad outputs") and not specifically aimed at you, and I didn't make that clear, so my bad.
Yes, this is literally what that means.
What exactly does "unguided" mean to you, then?
You might be surprised at how much benefit you get from the model council[0] pattern.
The common pattern I've seen is "build with Claude, review with GPT", or vice-versa.
It's been proven to be very effective.
0: https://www.google.com/search?q=model+council+pattern&udm=14
Design is too expensive, we do agile. QA too expensive, we fire all of them, and claim devops is the now, which allows us to fire the Ops team too, 100% ownership from deisng to ops on devs.
One person with an agent can replace all these teams. Yeah mo profits.
What is the point of working at a startup if you’re dealing with millions of lines of legacy code ? Isn’t the whole point of startups to create & innovate with a clean slate and modern tools?
I’ve never seen a startup with a multi million line legacy codebase.
>”Have you worked at many startups?”
In response to a question about a legacy codebase at a startup. That implies that they think whatever they are doing is common. And forking a multi million line codebase and heavily developing it isn’t common for startups.
OP might just be working at a small software company or for one that broke from a bigger one and is now "startup" like?
And if you're saying (based on your other comments) that a 6 month window is enough to create a legacy codebase...that indicates a serious lack of experience or understanding as to what a legacy codebase is, or why they exist.
Accept that other people may ascribe different meanings/interpretations to words than you, and that if your reading of their statement doesn't make sense to you, perhaps you are simply reading it wrong.
Trying to hold someone else to your definition of words suits what purpose exactly? Are you just trying to "win" ?
But who cares? The point is any codebase over a few years old with lots of customers and a big surface area has lots of code, much of it "legacy" from the standpoint of a guy in 2026.
15 people is a pretty small startup, what if this is a 50-person startup?
Doesn't seem like that much to me, depending on what you're building and the size of your team.
15 devs putting out 400k LOC a year into the same codebase is not normal at all. I’ve never seen anything close to that kind of rate of growth across that number of people.
And 50 engineers is not even remotely close to a small startup. I have worked at a startup that had 50 engineers after a few years, but it was a multi billion dollar unicorn.
This whole thread around loc is depressing. It speaks volumes of some peoples inexperience working on actual legacy code. Legacy code is not just age or size but that the technical foundation is dated in a fundamental way. A giant monolith running on a now defunk framework using a database only one guy in canada knows about.
Case and point in my day job. The org that owns XMM development does not know how to recover a physical bench that is bricked because everyone who knew how has left. So now they just use simulators…
Interestingly, AI figured out some of this pretty easily for me. But the org has the exact same AI as i do. At the same time another org is close to a year into a greenfield rewrite that has been developed via agentic swarms. Absolute trainwreck.
AI doesnt make bad engineers good. Anyone who says they are doing 4 eng work likely would be without ai too. Those that claim otherwise, are the bad engineers.
Yes of course there are words that have very clear meaning, and that's not the case here.
"startup" and "legacy" - used in the context of this discussion - are not words like "left" and "right" where everyone has a universal agreed-upon understanding of their meaning.
- Suggest a better approach that makes the AI say, “That’s much simpler. And you’re right. My original plan was over-engineered.”
A lot of my high costs is because I just throw Sol at everything. If I were more selective and brought in Luna or v4 Flash every once in a while, I think I'd be more like ~$400/month. That's why I'm not aligned with the notion that "tokens are subsidized so that's why people are using so much": its not that I'll have to adjust to using less, its just that I'd need to think before I prompt a bit and be more judicious. I could easily see my raw token counts doubling or tripling in the coming months. I don't think that will change as subsidization subsides; though maybe lab revenue will; intelligence per dollar is getting cheaper every week. Its solely a function of adaptation to process, which takes time.
The productivity gains per token are the single most asymmetrical thing I've ever seen in engineering. The engineers on our team are pretty effective with tokens; easily that 2x-4x output as you're seeing, spending $20-$200/day. Some of our security folks have also started contributing more-and-more code, and they're on the other side: they'll spend hundreds a day running in circles, eventually producing these +/-30k loc pull requests that take ages to get merged and are littered with issues. They weren't writing much code before, so arguably they're more productive by some multiplier greater than 1, but I think the drag on the rest of the team, and potential issues with what they produce, has overall created a net-negative situation. Inversely, some other company functions have produced a few one-off websites for things like sales processes, and those have been a huge win. The asymmetry is wild. There's almost a valley of incoming skill where if you know nothing about code, you'll leverage it well; if you know just a little bit, it makes you super dangerous; if you know a lot, you're the biggest winner. Really difficult situation to navigate.
In my experience at the begining of the full agentic coding loop in our company we were more hands on with the codebase and had better judgement over the plans. Now it is quite often that the inital plan after executed needs more refinement and that made the plan review somewhat obsolete for us.
(and mid-to-long-term, often also short-term end up cheaper than weaker models)
this might change soon if we are reaching a certain capability threshold
but right now that's still the case
unless you are working on throw-away trivial stuff where iteration speed and trying many speculative things might give you an edge
I have had some $3,000 token days - even without Fable. I don't see how this is sustainable.
My personal 20x plans get so much usage for so cheap. The consumer subsidies are crazy, but alas I can't use them for work.
It actually gives me quite an uncanny feeling, bulldozing over years of human optimization work with a newer, "perfect" design. Like bringing an AK-47 back to the middle ages.
it's so hard to square such totalizing statements with my day to day experience with fable and sol, (every possible, improvement, really?? they are NOT omniscient) arguing with them/my colleagues' agents that no they have slowed down the system 200x with their terrible change, doing string operations on millions of db rows, trying to get it to understand that I don't care that it's calling it a "cache" if a cache hit is slower than what we had before.
These agents do let you learn codebases quickly, and produce code way faster. I don't look at IDEs all that often. But literally multiple times every single day I catch them doing something stupid.
I don't think its impossible that we could get better performance from the agents. I know ive tried all sorts of workflows and skills, few of which seem to have much effect on the things the models struggle with. I think a big part of it is encoding enough context for large codebases, and providing it with all the tools it needs to make it successful, things to automatically check its work, etc. But that's not automatic, in fact its generally a terrible judge of what it needs or what its bad at
but the revolution is it doesn't take that long. in like 15 minutes you can chat with fable and get to the meat of whatever the issue is with repeated questioning. and then it does the solution for you. so it's not magic but it's still like a 100x speedup.
Opus and I wrote a parallelized test harness and labeled groundtruth in around 2 hours.
In 2022 that would've likely been all I did for a couple sprints
That sense that you did something better in a few days than you would have in a month 5 years ago. It's like buying a table saw for wood working.
One crazy thing I think about often is how there are so many correctness and testing harnesses that would have taken weeks to build in the past so we simply never would have. We'd just do our best then wait and see what comes to the surface. This is a huge part of what makes it possible to actually make better software with LLMs in my opinion. It isn't just 'LLM codes better than I ever could' (that's often untrue still) but 'LLM enables me to make assertions about the program to degrees that would have been absurdly impractical in the past'. It's huge
In this case the design was also AI generated, and there were limited wins to be found because the design was already superb.
But that's also why it's now easy to justify the cost of an Nvidia or Intel inference server with Kimi K3 locked and loaded :)
you should probably be doing this discussion work along with Fable 5. It will give good feedback if you're working on the correct things.
> Come up with QA plan and have Codex Computer Use execute on it
QA plan should be part of the above "design + plan", not after it. The implementer needs to be able to fully test before publishing a PR. This is true whether humans or agents are writing the code.
> Manually spot check the final result (usually a sizable diff, thousands of lines, complete feature E2E, etc)
unfortunately this is not really scalable with amount of code agents can produce, so you need independent (fresh context) agent reviewers to help. Ideally they only escalate to a human when really stuck.
> I probably spend like $80 a day at least
at a small startup you should be on the $200/month plan(s).
I'm in an infra team and for the last 2 weeks or so I've been trying to understand whether a particular workload will catch fire if a switch is flicked. I'm also new to the team so partly it is me wearing training wheels, familiarizing myself with the telemetry etc, but I will state that I'm not completely lousy at this stuff.
No model in my experience can do anything remotely comparable to the work "what happens to the workload if this switch is flicked" needs. They can't even design a reliable quick experiment to answer what cast should be applied to the binary trace_id in table A for the join to table B to work. They will happily do something idiotic and then conclude that the join does not work.
Are you at least conversational in the subject matter? You're gonna have a good time just by paying attention and adjusting your workflow. If you're getting a lot of back and forth with it, its asking a lot of planning type questions, stop, step back, rethink the whole feature, and start again from the beginning with everything more fleshed out.
If you are in a brand new field, there's no way to bridge that divide. The issue is you don't know what is good or bad, or whether what you have learned is good or bad. You're in a sports car and you don't know how to drive much less what's track and what's field.
You can spend a lot of effort getting good at prompting towards writing tests and E2E tests to at least verify your app does what you expect it to, regardless of experience.
I actually have no doubt that I could replace my Opus 5 Low/Medium subagent profiles with Grok 4.5/GLM 5.2/Deepseek v4 Flash and perf would probably be pretty similar.
On top of that - highly recommend adding accurate cost counters to your statusline. You can't improve what you don't measure! (Or even have any intuition about).
mate, if youre not using subscription then youre spending waaaaaaay more. the plan itself with fable/sol will most likely have already cost more then $80 -- ime thats more like 500-2k/day of usage.
most harnesses let you see the usage in the status bar, i encourage you to enable it
Possibly, but the output of a 2022 engineer is about 1/10th of the output of a 2010 engineer, so it's an extremely low bar.
I would love to see what these tools create but outside slop there's never: This works, is in production, here's the code.
Any day now.
It was all generated in the last 4 months, so don’t come at me with the “but did they use a model from the last 6 months” nonsense.
> I probably spend like $80 a day
This doesn’t sound like “unlimited”, I spend more than this out of pocket per day and I have a strict budget.
Wait what? I don't understand these numbers. I spend $1k/day
Your story about being told to use AI for everything I was expecting you to be well over that
no one’s AI spent will be in question working a single terminal with carefully planned out and executed process you do
On what planet do people start paying for things without keeping an eye on the costs and no-one notices until you have spent a crazy amount? I don't understand. You are either paying a fixed amount which you are happy about in-advance or you are PAYG in which case you would ballpark how much it costs.
Otherwise it reads a bit like a fake problem, because it didn't really happen, you just foresaw it (as you should) and added a few guide rails.
1. Insanely discounted starter plans. Claude $200/mo plan is like $5k-$8k of API rate usage.
2. Very limited cost visibility, they make it hard to figure out where you spent money (unless you're on the enterprise plan which is for people with unlimited money).
3. Nobody, not even the model provider, knows what your request will cost before it returns. You're writing a blank check every time you hit enter.
4. When you run out you run out very suddenly and disruptively. It's very hard to tell a developer on the 28th of the month "sorry, code by hand until the 1st of next month" so you tend to grant exceptions.
5. The price is changing all the time. New models come in, old models come out, prices change, caching behavior changes, harnesses change, etc. The cost of doing a single task is not predictable even if the task does not change.
6. Basically no volume discounting. Anthropic offered us 2% off for committing to $1M+ per year at API rates.
I manage AI spend for my team at work and I try really hard to keep costs under control but it's absolutely herding cats. Much harder than any other spending I've ever had to manage at work.
Happy to give you (or anyone here) some trial credits if interested! Email address in my profile.
The AI rollout has in large part been about giving management an opportunity to say "fuck you" to engineering.
Engineering payroll needed to be slashed and the C-Suite was happy to let their managers get some lashes in on the uppity workers.
The point of AI is:
1. It's bad engineering and everyone knows it.
2. Management is pushing it through anyway.
3. Deal with it, asshole.
The ROI is besides the point. It's about breaking the spirit of engineering labor with irrational decisions made outside of their control.
Edit: It is just as much about affirming the exalted status of management in comparison to labor. It's a bitter reinscription of the status hierarchy in tech; Management is good and labor is bad; Management is considerate and labor is ignorant; Management is strong and labor is weak; Management is superior and labor is inferior.
In addition, we're seeing people applying AI to more and more use cases, so token growth is very significant. Paired with consumption pricing, it's brought this problem to the forefront very quickly for lots of companies.
It is true that this problem can be mostly managed by the techniques we mention here. Those are actually pretty difficult to set up at scale, so many companies (including us) we only really did this in earnest once we started to see those large cost oscillations.
The main reason we shared this here is to maybe help other companies get infrastructure in place before massive cost swings rather than after.
What didn't happen is any analysis of cost/benefit up front. Many of the corporate decisions around AI have seemed characterized by companies blindly copying each other.
This is how AWS made its fortune.
Initially, you picked the Multi-ZA RDS db.t3.2xlarge instance because you figured "eh i have credits anyway". Two years later, someone looks at this and says "hey, this is expensive and I bet we can do everything we need on a machine half the size". But then they think "if i downsize it and that works, i'll get a thumbs up emoji on a slack thread. If i downsize it and it causes problems, i'll draw the ire of the whole team. I better leave it alone." And the truth is... by the time your company hits the end of those credits, you're probably at the point where that savings isn't gonna do much. Or maybe you are out of business.
And that is how almost every successful company that uses AWS eventually ends up paying six-figures or more annually.
They’re not saying they regret doing it, or that it was a mistake.
They’re just saying they’ve gained experience and have leveraged the tools to an extent their usage can be optimized.
Pretty standard business or life iteration.
If you have a 500k line codebase and even > 50% is written by agents, you are in a world of pain that won’t justify the costs longer term.
Now of course, there are products that just involve lots of code but are not actually complex. This is generally the project with like hundreds or thousands of features but most of the features are separate and don’t actually interact in complex ways. Think a task management app with hundreds of features like calendar, email integration etc. there I think agents gives you more bang for the buck. Just my thought, using agents at work.
Velocity and quality is through the roof, "10x" would be a regression at this point, the bottleneck is only figuring out what is relevant/good to deliver, the implementation is the output of the machine.
Have you ever listened to a politician talk for an hour and you can't remember what they talked about?
If some VC wanted to fund my crazy ideas for building personalized private versions of the internet because people are sick of the nightmare, I would vibe code the hell out of that; but at my current job I have existing customers to keep happy and make sure their financial data doesn't get lost or modified.
A bold claim to make with little to no supporting evidence
agree, this works, undervalued!
look at minimal agents that protect the context window:
- pi (https://github.com/earendil-works/pi)
- smol (https://github.com/smol-env/smol)
1) stronger models are more token efficient for open ended tasks because at the limit …
- stronger models can solve tasks that the weaker models can not solve
- stronger models make fewer mistakes, compose things better (cli, abstractions, …)
- navigate the code base better
- are better at removing and simplifying the code base again
2) dynamic request and task routing sounds great/obvious but is very very hard
- to benefit from caching you don't want to switch model or inference endpoint
- to _know_ a certain request can be routed to a weaker/cheaper model needs good context and a strong model to get right and often is still unknowable because the active coding session can go many ways and turn from trivial to challenging in a few turns, always in motion is the future, if you get it wrong you are back in the problem space of #1
for open ended coding sessions it is not so easy
that said: cheaper does not have to mean weaker, you want to look at the pareto frontier and stay up to date on new good models
there are many models like deepseek v4 flash and luna that are both cheaper and way better than most other models
I just spent 3 days reviewing the code of a vibe coding maximalist at work. They needed to touch 96 front-end files, ~50 back-end files. We now have `is_uk_company` spread throughout the entire codebase every time we need to output a date format. I would say 30% of the code can be replaced by a moderately more-wise approach of "you ain't gonna need this", 20% is actively polluting and/or harmful to the coherence of the feature.
I'm pretty sure this plan came from a braindump of a verbal meeting they had with our product team, using OpenAI's top models. There were 2 initial spreadsheets that were ABSOLUTELY INDECIPHERABLE, and now we have a new one labeled "FeatureX - Phase 2 to 4".
Just the date fields is a hilarious example of why I'm skeptical. Python (babel.date) and Javascript (the browser's locale) both have locale-aware date rendering. Using a library for country-based date and time parsing would have eliminated ~12 country-specific change sites... but instead we got `is_uk` and raw date format strings (`%d/%m/%y`) in-line inside already-enormous functions. Also, at a meta level, the backend code shouldn't usually be formatting the dates (data export may be an exception), we should be sending ISO dates and having the frontend format the code if anyone gives a crap.
I'm ~$500/day and I spent 3 days reviewing this stuff. Our CTO spent ~2 days this week reviewing this feature, I'm not sure what his daily rate would end up being. Just the date-formatting part cost me half a day... that's at least $250. A few hundred dollars on Xhigh tokens burned $2500-3500 in human review effort, and I was delayed from:
1. Getting an entire sentiment analysis feature deployed by helping our interns
2. Getting a load balancer solution working to automate a manual process
3. Improving our deterministic Cypress test suite
4. Getting a demo feature launched for our sales team that would bring in more money
There’s not even a distant comparison US lab rn which is sad
I would bet my entire Polymarket balance ($0) that some military contractors have already asked AIs on the public Internet to design software for them.
That said, this type of post needs to be read with product marketing context in mind.
> I think the most important aspect from it is the auto-routing
On the contrary, in white paper studies auto-routing is shown to destroy the single largest token cost they found, curiously not shown in their opening graphic. As they put it in their own words near the end:
"simple tuning of… caching settings… 50% reduction in … costs, with no observed quality degradation…"
They also aren't showing the harness called ‘pi’ which generally tops results (not only with ‘open’ models), and is consistent with Anthropic's recent "works better when we don't stuff 100k system prompts into context" about Opus 5. Recent models do better with less jank trying to prescribe behavior.
So then one wonders: why is this post written to say you should use a meta router but no mention that's offset by the cache busting, and show developers cost savings but no mention that Anthropic and OpenAI both have developer-facing utilization dashes already?
Perhaps it's so engineers can show enterprise procurement why they want to buy exactly what Databricks happens to have for sale.
1. Codex, Claude and others try to switch models being used at their level itself to manage the cost and outcomes
2. Now company like data bricks develops one more layer on the top of it to do the same task, of finding the base harness and applicable model
Companies like Codex and Claude are focussing/investing heavily on to ensure that people are using their harness directly or instead use APIs. Unless Databricks has some agreement in place they are violating the TOS and openly publishing an article about it. Would be interesting if openAi or Anthropic come back and claim for the API usage prices and all the savings go away.
[0]: https://support.claude.com/en/articles/9797531-what-is-the-e...
[1]: https://support.claude.com/en/articles/9266767-what-is-the-t...
Anyone using the enterprise plan are charged the API pricing, however the article is not clear if Databricks is using enterprise plan or not which is why added the following disclaimer
> Unless Databricks has some agreement in place
June 15 changes would be the keyword to check, but at least in anthropics case it's... Complicated
One more datapoint for the thesis that OpenAI and anthropic aren’t viable, sustainable businesses, and cannot justify their $1T valuation and the level of compute commitment (reminder that OpenAI committed to >$750B in infra spending for 2030)
1. Thinking about how to optimize AI spend is one more thing to think about when I have too many things to think about already;
2. If there’s even a minuscule chance that me selecting down the model results in a task getting delayed because for that particular thing the model happened to be “too dumb”, I’m just not going to do that. Because I’m evaluated for what I ship and saying, “oh, not happening this week because I downgraded to Haiku for the exploration phase, and that resulted in a faulty plan, so it slipped a day” just doesn’t cut it.
3. Why should I spend time thinking about costs, and risk being slower because of that, if the next guy doesn’t?
Kudos to databricks, I also find it interesting that such different companies (Stripe, Ramp, Databricks) are all building the exact same internal tools.
I think building companies is going to look more generic in the future because intelligence is an API now.
Have you tried measuring Gemini? now that you have the router it should be a simple task. Thanks!
Without the insight into agent performance, any changes like this feel like a gamble to save $$ at the cost of developer productivity
I'm actually working on building generic repo-specific benchmarks at https://stet.sh ;)
In large user populations like at Databricks I think the ultimate answer will come from experimentation instead of offline evals. We are already doing this in small groups, exposing them to new candidate models and then measuring per-developer cost and perceived quality changes.
If my company told me yeah we’ve decided you don’t get Fable or Opus 5 because it’s too pricey, you gotta use GLM whatever, I’d be displeased.
My workflow is very simple:
1. develop requirements for code change
2. take manual notes for implementation, maybe use LLM for some discovery/investigation
3. present notes to frontier LLM
4. develop implementation plan (bulk of work)
5. let LLM rip
6. review diff, manually fixing/refactoring code as necessary, sometimes prompting for revisions
7. get automated LLM review
8. get human review
this reliably produces the work of 2-3 pre-AI senior engineers with a lower bug rate, equivalent performance, robust edge-case consideration, etc.
Does the LLM produce over-engineered solutions? All the time. I stop it from doing that, or manually fix it myself.
Does the LLM always adhere to the best system design? No, not at all. I often have to guide its design into a better, north-star aligned one.
I don't just sit in front of my terminal and say, "Ok Claude, build the app." It is a very iterative process, and not without its potential pitfalls.
But it is very, very productive.
I think the more interesting lever is the fourth they mention: token efficiency.
> By the time costly LLM inference occurs, the user's initial statement accounts for only a negligible fraction of the data fed into the AI system, meaning costs are dominated by context the user did not explicitly include.
I think there’s still lots of low hanging fruit in regards to monitoring and improving agent work. Look at your sessions. Look at how much time and context is being spent on, say, a web search returning dozens of results when one good single-pager doc would’ve been better.
But the question remains, AI hasn't shown any explosion in growth/revenue for most companies. The effects seem muted.
Similar idea re using the same UI across all models, but the agent can modify the harness config as well as start/schedule sessions, etc. The Kanban board feature can be used to orchestrate agent driven workflows, and the agent itself can modify the Kanban lanes. Basically the agent can do all the same things you use the UI for.
Example prompt: “schedule a session using Opus 5 with max thinking for every feature in the PRD on the canvas. For every session make the prompt instruct the agent to review end to end test coverage for the feature and create a report on the canvas for test gaps that you find. Schedule the sessions overnight tonight spread evenly throughout the night, and have each session set to retry upon token exhaustion.”
I think it’s more meta than other meta-harnesses, but I’m biased because it’s my pet project.
I haven't gotten a chance to test the multi-agent capabilities, but the DeepSeek Flash prices are so low that I probably will soon.
I feel like I have an advantage over big companies, if I can use the best models on a subscription and not worry about costs much, when they can't do the same as outlined in the article.
Have you all attempted finetuning smaller OSS models on your repos for coding?
Maintaining the cache is extremely, extremely important, so we're iterating fast but that's a major factor we track in the router's development. Couple things I'd look at:
1. The cache is generally reset after a compaction - this is the best time to make a switch if you want.
2. In many cases, the max duration of a cache is 1h, so if a session is being resumed after a long time, that's also a good time to re-assess the complexity.
We're iterating fast here and learning a lot! Definitely a lot to think about it in this area.
https://www.databricks.com/blog/introducing-omnigent-meta-ha...
We're calling the layer coordinating harnesses "meta-harness'
An AI-edited post by the way.