And I'm seeing almost no self-awareness from leaders. They are making decisions about things that they just don't understand. And are completely unworried about it. Just blindly following whatever the news cycle is about AI.
As I was saying, you're all fired.
I predict that will be their comeuppance; it will begin a new era in history.
What he says he's consistently hearing from them mirrors what I saw at my own employer: they thought they had ROI metrics, but they actually only had usage metrics such as "lines of code committed" or "number of pull requests". The only way those could possibly work as an ROI measure is if your business charges customers by the line of code.
Eventually I got tired of it and got back to work.
Which reflects another thing I’ve seen at work. A lot of what AI coding has enabled is diving headfirst into quagmires. Our costs have spiked - not just because of the token spend, also because we gotta pay the cloud platform to run all these new services, operators to operate them, marketers to market them, etc. - but revenue hasn’t budged.
Why be a normal guy that waits to see what happens and is measured and pragmatic when you can get attention basically through the whole cycle by being the earliest adopter, adopt it to the maxx, then also be the loudest big brain when the tide changes and be praised for "taking hard decisions" when you revert everything you said so far?
The fakemaxxing economy.
As a leader, pushing for rapid change cannot really be nuanced lest the push dissipates into the organization's entropy.
It's irrational to push for tokenmaxxing (literally "please increase our AI spending") and not expect that this is the result you are going to get. You won't get productivity increase, since that is not what you are pushing for - you will get token usage maximization (engineers running inane agentic tasks against your code base to increase usage, using company paid AI for their side projects, etc, etc).
Perhaps that is what they were trying to do, but the reality is that all they will have got is a large token bill. The decision makers may have hoped that tokens would be used in most productive fashion possible so they could evaluate if the cost was worth it, but what they will have actually got is what they asked for and measured, high token usage (applied to whatever people needed to do to get their usage stats up, regardless of productivity).
The other business-as-usual factor is that there will be false reporting up the chain, so if the company understands the CEO want to see high AI usage and productivity gains, then s/he will see high AI usage (a large token bill) and will be fed success reports of corresponding productivity gains.
In a typical corporate environment, if all your peers are reporting success, achieving what the CEO wants, do you want to be the only one reporting failure? So - everyone reports high AI usage (easy for the employees to make happen), and most everyone also reports productivity gains if they understand this is the expectation.
If you want to report productivity gains or cost savings from some initiative (increased AI usage or whatever) and need some stats to point to, then you just point to whatever is working, for whatever reason, and attribute the success to the new initiative.
In a company I used to work for, one manager, when pushed to increase machine learning usage (a few years back, before ML became AI), just renamed his product from foo to foo-ML (with ZERO ML usage), and reported how well it is working. He has since been promoted twice.
Do these measurements have sufficient information? As much as any, I'd guess. It sounds like you already know that it's pretty hard in general to measure the productive output of software development organizations.
Don’t play their game and call them leaders. They are management, bosses, executives.
> They are making decisions about things that they just don't understand. And are completely unworried about it.
Clowns, even.
> Just blindly following whatever the news cycle is about AI.
But followers might be most apt.
——
This is such a huge pet peeve of mine. Describing management goofs using their language that makes them sound all-so-brilliant. We constantly watch these people do the dumbest shit and then they go around describing themselves as “thought leaders” and “servant leaders”. When, really, most are just clowns with fragile egos.
And, while I’m rambling, they’ve tried to take away the fact we are workers by calling us individual contributors. Using language to attempt and hide the hierarchy and power dynamic at play. It just…bothers me so much.
And many of them still claim they are "risk takers", but have effectively insulated themselves from risk by socializing losses.
You're falling into a common trap here: the ambiguity of the English language.
"Leader" means multiple different things. Yes, it means someone who has leadership qualities—who genuinely inspires those around them to do better, or who boldly marches into the unknown and gets people to follow them.
But it also means "someone in charge of a thing."
Now it's certainly true that many people in charge of things who are also really bad at actually inspiring or getting people to follow them (aside from with threats of destitution) also play on that ambiguity to try to convince people that because they're in charge of things, they must also be Good Leaders, and that's crappy...but yelling at others for using the term casually is very much an "old man yells at cloud" situation.
I once worked in a company that had soviet-level efforts to push LLMs into everything, someone eventually made the classic "Natural Language -> SQL Query -> Magic Result in webpage" and got promoted, the tool got mandated for every non-tech employee as part of an AI-boosting effort (people pushing metrics up).
One day I wake up with a product person in despair because the tool couldn't handle what looked like a very simple aggregation, I stopped what I was doing, crafted a 30-line SQL query over HORRIBLE TABLES, a couple CTEs and window functions here and there got him what he wanted. I found out later that single query that took 30 minutes to make saved him from inheriting a 6-month effort to create a microservice dedicated to patching said tool.
Understanding this was one of the most important things in my career.
It would only be laughable if they waited way too long to reverse course, but I don't think that's the case.
Over/undershoots and corrections are of course unavoidable and normal; the absurdity is at the magnitude and rate of change. Furthermore, this is giving it the benefit of the doubt, that measuring AI spend is a good indicator; that's arguably also in dispute. To stretch my car analogy a bit more: it would be like the cruse control system has to hit the target speed, but it only has data from the O2 sensors.
[0] I know that the "classic" cruise control system cannot apply the brakes, but hey no analogy's perfect.
How much that makes it into enterprise pricing is TBD, since none of the hyper scalers are making money yet of selling AI inference.
Almost all businesses are ahead of the gun. For most of their use cases, AI is either not yet good enough on its own, or good enough but too expensive.
No one wants to get left behind, so everyone's trying to get onto it now, even though it's not ready for what most enterprises want to do with it.
It's easy for them to look at a small startup without billions of lines of legacy business logic debt and see them having success and wonder why they can't have just as much - or more - why they're bigger so they should have better and more success, right???
Wrong...
But when it gets ~99% cheaper for local inference over the next 4 years, at the same time the price per watt improve 4x -> a lot of those cases will start to pencil out.
Historic trends, every 18 months, performance for the same level of quality has gone down 90%.
See: https://www.reddit.com/r/LocalLLaMA/comments/1gpr2p4/llms_co...
And Chart 13 here: https://www.rdworldonline.com/ais-great-compression-20-chart...
And here: https://epoch.ai/data-insights/llm-inference-price-trends
Historically, algorithmic gains are only ~30% of the pie, but there's enough out there to get to 10x, with just what's available already. The other ~70% of the pie is better training data (often synthetic) and distilling frontier knowledge. There's no sign we are tapped out on that front.
Additionally, GRAM (from ~10 days ago) is likely to be a 5-10x on its own (if not substantially more for smaller models). It's unlikely within 4 years LeCun's JEPA ideas and similar ideas like GRAM applied to LLMs have ZERO impact. The preliminary results are absolutely astounding (5000x better reasoning - this is not peanuts).
Further, that's not even counting that cost per watt is still dropping ~2x every 2 years on its own on the hardware front.
If you look at the "cost" of inference. People think it's electricity - but it's currently almost ~80% hardware amortization. The memory shortage is not going to last, nor are Nvidia's ~80-90% margins.
The human brain is still 8-10 orders of magnitude more efficient than the best LLMs of today. With ~1/10th of global capex riding on AI, if you don't think they're going to knock of 2 orders of magnitude more, when it's this obvious and easy... I don't know what to tell you...
Sure, it might take 6 years instead of 4. My crystal ball isn't perfect.
See https://arxiv.org/abs/2604.04364.
This won't really show up in benchmarks, but it will impact real world usage on the most common use cases.
I'm doing a study right now on the impacts of better context for small models to fix bugs.
A very dumb algorithm can make small models perform at 10x+ model sizes. I'll be surprised if it can't get to 20x+
Thank you for sharing this and for having the intellectual courage to hold to a sound reasoning that may be unpopular initially.
I think what will also happen, once we get past this current CEO AI FOMO mania, is that companies will start to look at AI spending more rationally like any other company expense, and will revert to more rational decision making.
Even if the cost comes down considerably over the next few years, that's plenty of time for companies to look at their financial results and question why AI expenditure isn't resulting in increase in revenue and/or profitability.
Do you mean the marginal cost by the producer, or the cost on the consumer? I can't see the price of electricity falling much, and the demand curve is apparently exponential if the hype is to be believed.
Computing has always been about how to wring out more efficiency. The ENIAC was 150,000 watts, with 3 phase 240 volt power, and cost about $500,000.
My day to day laptop (a year old) is 35 watts, with 1 phase 20 volt power, and cost $1,000, so that's 99.98% less power consumption, 99.8% cheaper, and it has about 10 orders of magnitude more computing power, all on a time span of 80 years.
There's always a chance we'll have some dramatic gains far larger than DeepSeek's optimizations a year ago, but it hasn't happened again yet at even that scale. It would be nice but I certainly wouldn't count on it.
People are willing to pay more for BETTER quality.
You obviously haven't seen DeepSeek v4 Pro's pricing if you think pricing only goes up...
Let's look at GPU prices as an example. Around 12 years ago, I bought a GTX 970 for around $350. That was considered a very good GPU at the time. Today, the "equivalent" GPU model (RTX 5070) now costs almost double. Of course, the newer GPU is much more powerful (more than double, in fact), but all the things you'd use a GPU for have also advanced and now expect an entirely new level of performance as a baseline, such that the older GPU is fairly worthless today. So most people agree that GPUs in general have become more expensive.
Regarding DeepSeek's price: it's obviously subsidized, and unlikely to match the actual inference cost right now.
Then buy $10 (or $2, if you're cheap, and they take PayPal) of DeepSeek credits.
Whilst you're at it spring for a Claude subscription too and GPT.
Switch models between Qwen, DeepSeek Flash, DeepSeek Pro, and you can meet 99% of your code generation needs.
Hop over to Opus 4.7 (or 4.8, but I haven't really used it yet) and GPT-5.5 when doing very complex architecture/design or troubleshooting something where DeepSeek Pro is getting stuck.
It is ridiculous how cheap this stuff is now. It's affordable at third world prices.
> spring for a Claude subscription too and GPT.
You started with some random pricing then veered off into impractical hand waving. Far above third world prices...unless you count the USA as third world, I guess.
If you have the $$, do the extra stuff. People who like to play video games often have a very fancy graphics card that sits idle during their work day.
And the technology already exists on the algorithmic front TODAY to lock in another 10x gain -> when, typically, algorithmic gains only account for ~30% of that drop and the other ~70% comes from better data (often synthetic) and knowledge distilation from frontier models.
Just look at DeepSeek's pricing...
The Chinese, since they lack computing hardware due to US export controls, are.
So I just started trying CodeWhale (https://github.com/Hmbown/CodeWhale) with DeepSeek V4. I expected to be impressed by the abilities (which still require plenty of oversight). I didn't expect to be completely shocked by how cheep it is. After most of a week of using it 4-8 hours a day, which would amount to a full week of coding in many jobs after you account for non-coding activities, I'm about to hit $3 in total usage. So we're talking $10-20 per month for single-agent use by a full time software developer? And I'm sure some of my usage is waste as I'm still getting my head around things like compaction. If I take a break for a few weeks, I pay nothing because there is no subscription.
If DeepSeek and Xiaomi MiMo stay within a few months of the US-based models in terms of capabilities and US companies don't figure out how to drastically cut prices, I can't see how China hasn't already won. Protectionism would be one reason, but that might be ceding 50-90% of the total addressable market, and bring us closer to moving knowledge work out of the US the same way we did with manufacturing because it's too expensive in the US.
The first big task was to find the common bits and abstract them out. It did a great job of creating a plan, summarized in a table, that gave a name to shared chunks, the line numbers in various files where they appeared, line counts of new functions vs. removed bits, and some pros/cons about splitting out each chunk. It was very well "thought out", so I told it to go ahead. It did a nice job other than straying from my coding conventions. That gave me a chance to build out my AGENTS.md file (it helped with that, too).
Once that was done, I had it create automated tests for the newly abstracted parts. I think this is probably a bad practice... I believe humans should at least define what the tests are testing so that there is a deeper understanding of what oversight is in place. But I was just trying things. It surprised me how well it did. The biggest surprise was that the tests seemed quite inspired by vision. It would try different parameters and then have comments about making sure the shape protruded in a certain way, then code that did that. I expected it to refactor a bunch of the code to make it more testable. It found a way to not touch the code while testing everything I asked it to with just two simple mocks - I hadn't foreseen that, but it felt quite practical. It was passing around several opaque tuples in the tests and accessing items in them by index. I prompted it to replace the first one with a frozen, kw-only dataclass. Then a second. On the second request, it saw the pattern and did the rest without me asking. It created 44 tests across a handful of files.
The next part is where I was the least happy. I use ruff and ty to check my code with almost all checks enabled. It was mostly good about the ruff issues. But for the type checking, it just wanted to disable 6-8 rules for the entire repo in pyproject.toml, or at least for all the tests. I had to repeatedly tell it not to and it kept telling me it wasn't recommended. When it finally gave in, it fixed most of the type issues (build123d has lots of types specified, but many operations result in type conflicts because things are so deeply overloaded). The things it didn't fix, it just left a comment to ignore type checking altogether on that line. After I did a little more brow beating, it finally changed the comments to only disable specific rules. To be fair, and unlike most of my other repos, I've had to spend way too much time getting types right in this repo myself.
My last task involved a small library management system for our little town library (tracking library cards, books, DVDs, check-outs/check-ins, etc.). I inherited it from someone who had built the entire web app out of bash/awk/troff scripts with the data in text files burdened by a lot of schema changes that he didn't really know how to deal with. I'm halfway through moving it to Python/FastAPI/SQLite. I asked it to do a security audit of the entire code base, both the newer parts and the old parts that are still in bash/awk/troff. It found everything I knew about and a few things I didn't know about. It made a decent assessment of the risks/impact of each issue. It also called out design decisions that were good security practices. One of the next big tasks will be to see how it does at continuing the migration - it has enough examples of how I've done it that I suspect it can do something fairly consistent with my thinking. I'll probably have it do one or two web pages. When I feel like it understands what I'm after, I'll tell it to use sub-agents to do the rest. I'll be very happy if I don't have to tease apart any more troff scripts that are generating PDF files!
I doubt it is really any different to what the US labs do [1]. I never really bought the "they were basically all just distilling from us" shtick from Anthropic, I just assumed they were either comparing or also creating training data as basically any lab is doing.
[1]: https://www.reddit.com/r/ClaudeCode/comments/1tqaist/opus_48...
And fusion power is just 2 decades into the future!
We have little visibility into current frontier model costs at mass scale. As a broad historical trend, tech costs tend to fall over longer time periods but your claim far exceeds Moore's Law rates in its heyday - and that heyday is long gone.
In 2021 TSMC announced it was increasing it's price per gate for new nodes for the first time in its history. In the past five years cutting edge nodes have delivered ~8-15% real-world performance gains on average at costs at least 10-20% more than the last node. If you're positing a string of unprecedented efficiency breakthroughs in LLM algorithms - such extraordinary claims require extraordinary evidence.