Ricardo Arguello

Meta shutting Claudeonomics down two days after this scoop is the cleanest evidence that the leaderboard itself was the problem, not the people on it. Ranking smart engineers by consumption produces consumption. The line Jatin Garg adds further down the thread, "the leaderboard is the tell," names it well. Eric Ries published the structural-fix argument the same week in Foundr (Incorruptible): the fix isn't behavioral, it's redesigning the architecture that allows the behavior. Cross-read: iqsource.ai/en/blog/addict…

Exclusive: Meta employees are “tokenmaxxing” and competing on an internal leaderboard called “Claudeonomics” for status as a token legend. Over a recent 30-day period, total usage on the dashboard topped 60 trillion tokens.

The "smart people hitting targets they assume leadership wants" line is the cleanest description of how Claudeonomics happened without anyone deciding for it to happen. Same pattern at Uber and Amazon. The fix can't be at the engineer layer because the incentive itself is the problem. Eric Ries published the structural-fix frame this same week. Cross-read: iqsource.ai/en/blog/addict…

Token usage is part of perf evaluations at Meta. This is just smart people (Meta only hires smart folks) hitting targets they assume leadership wants them to hit so they get that exceeds expectations (or above) rating + avoid below expectations (Perhaps this is Meta’s goal btw)

The scoop tells you the engineers preferred Claude Code and finance pulled it anyway. Which means it wasn't a quality call. It was a structural-economics call that nobody designed for at deploy time. The interesting thing is that Microsoft, with the deepest pockets in the room, hit the wall first. Wrote about what designing for it looks like, using Eric Ries's Incorruptible frame: iqsource.ai/en/blog/addict…

scoop: Microsoft is starting to cancel Claude Code licenses. Engineers in Microsoft's Experiences + Devices team will have to transition to GitHub Copilot CLI by the end of June. Details in my Notepad 📒 issue, live now for subscribers 👇 theverge.com/tech/930447/mi…

The $180M/month number is the cleanest evidence yet that the problem isn't pricing, it's incentive design. Nobody approved $180M for a leaderboard. Meta's Claudeonomics produced it because the incentive rewarded consumption and not net result. Same pattern at Uber and Amazon. Eric Ries this same week named the fix: structural design before deploy, not behavioral pleading after. Cross-read: iqsource.ai/en/blog/addict…

60 trillion tokens in 30 days. The math on Meta's "Claudeonomics" leaderboard is one of the wildest things I've seen this year. At Anthropic's Sonnet pricing ($3 per million input tokens, $15 output), even a conservative blended rate puts that north of $180 million a month. On one vendor's API. The top individual user averaged 281 billion tokens. Run that at Sonnet rates and you get somewhere between $843K and $4.2M per month in token spend for a single employee. That person is consuming more compute than most Series B startups burn in a year. Meta has roughly 79,000 employees. Divide 60 trillion tokens across that headcount and you get ~759 million tokens per person per month. But usage follows a power law. The leaderboard exists because a small fraction of engineers are consuming a wildly disproportionate share. The top 1% are probably burning 50x the median. Here's the part that should make every AI company pay attention. The leaderboard is called "Claudeonomics." Meta, a company that built Llama and has its own foundation models, is running an internal status game named after Anthropic's product. That tells you everything about where the actual coding workflow loyalty sits right now. Jensen Huang proposed giving engineers a $250K annual token budget as compensation. Meta is tying performance reviews to AI usage. One Swedish engineer's employer reportedly spends more on his Claude Code tokens than his salary. We're watching compute spend per engineer approach and exceed the cost of the engineer. When the token bill is larger than the payroll line, the math on headcount changes permanently.

The "subsidy ends" framing is the right read on the price side. The structural read on the demand side is the one most coverage skips. Microsoft's engineers preferred Claude Code, which means usage was real, which means finance wasn't fighting hype, finance was fighting an addiction-architecture pattern that nobody designed against. Wrote about what design-against looks like (Eric Ries's structural-fix frame applied to consumption): iqsource.ai/en/blog/addict…

The more interesting part is that Microsoft’s own engineers liked Claude code best…and they’re cutting it anyway Token pricing is making enterprise actually look at what these models cost to run and that could be a problem for the labs when the subsidy ends. Lots of cheaper options out there and not just Chinese open source anymore

The settings.json audit cuts the per-engineer bill. The harder fix is the one above it. If finance still doesn't see which processes earn back their tokens and which don't, the next pilot just rebuilds Claudeonomics with cleaner config. Wrote about why the structural contract has to come before the tokenmaxxing config: iqsource.ai/en/blog/addict…

Claude Code Head invented a word for what most config files aren't doing: tokenmaxxing. 90% of it lives in one place: ~/.claude/settings.json. 125 keys in v2.1.105. official docs cover 40. 4 of mine aren't anywhere in the docs. i audited the whole file - 18 settings actually decide what your month costs. > enabledPlugins - mine had 14 on, 4 actually used. each idle plugin loads hooks and SKILL.md at session start. > mcpServers.enabled - false instead of delete. 9 unused servers cost ~30K tokens of schema every session. > cache_control breakpoint at the static/dynamic boundary moved my Opus month from $340 to $87. tokenmaxxing isn't a vibe. it's a file with 125 keys most people never open. watch this tonight instead of the new Rick and Morty drop.

The Naga and Catanzaro numbers are the real story. Nobody at Uber decided to burn their annual budget in four months, and Catanzaro didn't decide compute should outpace payroll. Both fell out of incentive structures that nobody redesigned when the billing model changed from flat to per-token. Eric Ries this same week argued the fix is structural, not behavioral. Read the cross-read: iqsource.ai/en/blog/addict…

Microsoft just banned its own engineers from using AI. The tool was literally costing MORE than the humans it was supposed to replace. They lied to you about AI adoption and now the whole narrative is blowing up: Microsoft gave thousands of engineers access to Claude Code six months ago and encouraged them to use it. Engineers loved it and adoption exploded. But then the invoices arrived. Token-based pricing means every query, every code review, every debugging session costs money. At scale across 100,000 engineers, the numbers became so large that Microsoft issued an internal order to cancel nearly all Claude Code licenses by end of June and force everyone onto their own cheaper tool instead. The company that invested $5 billion in Anthropic just told its own people to stop using Anthropic's product because it costs too much. Uber's story is even worse... Their CTO Praveen Neppalli Naga told The Information that the budget he planned for the full year was "blown away already" by April. Uber had rolled out Claude Code in December 2025. By March, 84% of their 5,000 engineers were using it with 70% of all committed code coming from AI systems. Heavy users were burning $500 to $2,000 per month each. Naga himself spent $1,200 in a single two-hour demo session. The company had even built internal leaderboards ranking engineers by how much AI they used. They literally gamified the spending and then ran out of money. Now look at what Nvidia's own VP of applied deep learning Bryan Catanzaro said to Axios last month. Direct quote: "For my team, the cost of compute is far beyond the costs of the employees." This is a VP at the company that SELLS the chips saying that using AI is more expensive than paying humans. Think about what this means for the entire AI narrative. Every CEO on every earnings call for the past two years has said the same thing: AI will make us more efficient, reduce headcount, and cut costs. The stock market rewarded every company that said it. Fired workers, stock goes up. Announced AI adoption, stock goes up. But the actual companies deploying AI at scale are discovering the math doesn't work. The MORE employees use AI, the HIGHER the bill. Goldman Sachs forecasts a 24x increase in token consumption by 2030 as companies adopt AI agents. Gartner just published a report showing that even though individual token prices will drop 90% by 2030, total enterprise AI costs will go UP because agents consume exponentially more tokens per task than basic tools. Meta built an internal dashboard called "Claudeonomics" to track which employees use the most AI. Amazon started pushing engineers to "tokenmaxx," their internal term for consuming as many AI tokens as possible. Both companies are spending hundreds of billions on AI infrastructure this year alone. And Microsoft, the company that bet its entire future on AI, just told 100,000 engineers to stop using the tool they liked best because the per-token bills got out of control. The companies building AI are telling investors it saves money. The companies using AI are finding out it costs more than the humans it was supposed to replace. And even the company that makes the chips just admitted it through its own VP. This is the gap nobody on Wall Street is pricing in. $725 billion in AI infrastructure spending this year across Big Tech. And the first companies to actually deploy these tools at scale are already pulling back because the economics don't work. What do you think?

The architecture-of-addiction frame is correct and the "it IS the business model" close is where it breaks. Eric Ries published the counter-argument the same week in Foundr: structural problems get structural fixes (mission-locked charters, perpetual purpose trusts, Novo Nordisk foundations). Applied to AI, the fix is a selectivity contract designed before deploy, not a "use it carefully" memo after. Wrote up the cross-read here: iqsource.ai/en/blog/addict…

Marc Andreessen admitted on Joe Rogan that AI is making people less efficient. The guy who funds half the AI industry. On a podcast. Just casually dropping it. Same week: Nvidia's VP said compute now costs more than his employees. Microsoft canceled 100,000 Claude Code licenses because finance couldn't stomach the bill. Uber burned $3.4 billion in AI budget by April. And here's the detail everyone's glossing over: Uber didn't just adopt AI. They gamified it. Internal leaderboards ranking teams by usage. They made burning tokens a competition. A sport. It worked. Adoption went from 32% to 84%. Engineers loved it. They used it for everything. They stopped thinking about whether a task needed AI. They just used it. For everything. Always. And that's when the budget died. The tool was so good that people stopped being selective about when to use it. And the moment you stop being selective, the cost goes exponential. Because token-based pricing means every thoughtless query costs real money. This is the part nobody wants to name: AI doesn't have a cost problem. It has an addiction architecture. Flat-rate software trained an entire generation to use tools without thinking about cost. Now AI billing is per-use. But the habit of "just use it for everything" didn't update with the billing model. Uber built a leaderboard that rewarded maximum consumption of a product billed per unit consumed. Then acted surprised when the bill arrived. Microsoft's engineers unanimously wanted to keep Claude Code. Finance killed it. The people using the tool said it was the best thing that ever happened to their workflow. The people paying for the tool said they couldn't afford how much the users loved it. We built something so useful that the only way to sustain it is to stop people from using it freely. And that contradiction isn't a bug in the business model. It IS the business model. It's how every AI company makes money: build something addictive, bill by consumption, and wait.

Prompter dice que la adicción a IA por tokens es inevitable porque ES el modelo de negocio. Eric Ries publicó la respuesta esta misma semana. Los problemas estructurales no se piden, se rediseñan. La compuerta Go/No-Go es eso aplicado a consumo de IA. iqsource.ai/blog/adiccion-… @IqSource1445 @alex_prompter @Mnilax

@LajuaneTorrey @EytanStarkman @Starbucks @Starbucks_cr Exactly!

@EytanStarkman @Starbucks @Starbucks_cr @RicardoIQSource Starbucks had a data infrastructure problem disguised as an AI problem. Inventory systems manually updated at store level can't feed an intelligent model reliably. You can't layer intelligence on top of inconsistent inputs—the model just automates the chaos.

@Starbucks didn't fail at AI. It failed at strategy. You can't buy an app off the shelf and staple it to 11,000 stores. That's not deployment. That's wishful thinking. @Starbucks_cr , talk to @RicardoIQSource about AI Maestro. Costa Rica sandbox. Weeks to prove it. Then scale. This is solvable. Just not the way $SBUX tried. cnbc.com/2026/05/21/sta…

Glasswing just gave your framer vs frame thesis a hard data point: 10K vulnerabilities found in a month, 75 patches applied. Several open-source maintainers asked Anthropic to slow disclosures. The 75/530 ratio is the framer scarcity quantified. Borrowed your framing to argue what CTOs should do this week before deploying any AI finder: iqsource.ai/en/blog/where-…

We’ve automated every single thing we can @every with AI agents. And yet there’s way more human work to do than ever. We’ve gone from 4 -> 30 human employees since GPT-3. I wrote a report on the structural reasons: how AI makes expert competence cheap, why that drives up demand for experts, and why the dynamic only intensifies as we approach AGI. After Automation: every.to/p/after-automa…

The Jevons frame applies cleanly here. I have watched this exact pattern twice before as a programmer since 1990. Compilers commoditized assembly and architects exploded in demand. Frameworks commoditized boilerplate and senior engineers exploded in demand. Now AI commoditizes implementation and the framer role is what becomes scarce. Wrote it up with the Glasswing data: iqsource.ai/en/blog/where-…

This is a fantastic post about why jobs aren’t going away in the way some predict. We are constantly making the mistake of confusing task completion with AI with being able to eliminate the whole job. Even as we can automate one or many tasks within a job, the definition of the job almost inevitably just expands to do vastly more of those tasks, do them at a higher quality, or move on to the type of task that hasn’t been automated yet. And as a result of being able to do more of the tasks or at a higher quality level, the job becomes valuable in a new way. And in many cases for now an entirely new audience as well. This will be true for coding, legal work, sales, or marketing. The small business or non-tech company that wants to now take on larger software projects finally can, and they’ll hire to do so. The small business that couldn’t afford a full marketing agency can hire or contract out to a marketer that can do as much as an agency did before now with agents. And so on. Don’t fall into the trap of confusing tasks with jobs.

Exactly. The 75/530 ratio in the Glasswing report is the human triage bottleneck made visible. The maintainers asking Anthropic to slow disclosures are the same signal from the other side. Wrote about the three concrete moves a CTO can make this week to inventory framer capacity before deploying any finding agent: iqsource.ai/en/blog/where-…

Here’s a key line in this mythos update. This is precisely an example of why engineers don’t go away, ever. We’ve made it far easier to create and find security issues, which means the new bottleneck is our ability to actually review, respond to, and fix the issues. Far from AI magically solving all of this, there still is major triage work and human judgment required to do the follow on work to actually protect systems. As a result, we’re about to enter a security engineer boom. Jevons paradox all over again.

The number that matters here is buried later in the update: 75 patches applied out of 530 disclosed, and several maintainers asking you to slow down. That ratio is the framer scarcity made visible. Wrote about what it means for enterprise AI deployments and how to inventory triage capacity before deploying any agent: iqsource.ai/en/blog/where-…

Last month we launched Project Glasswing, our collaborative AI cybersecurity initiative. Since then, we and our partners have found more than ten thousand high- or critical-severity vulnerabilities in essential software.

10,000 vulnerabilidades críticas encontradas en un mes. 75 parches aplicados. Eso es lo que Anthropic publicó esta semana sobre Glasswing. Y los mantenedores le pidieron que bajara el ritmo. El cuello de botella se movió. Y ningún agente lo va a romper por ti. iqsource.ai/blog/cuello-de… @IqSource1445 @AnthropicAI

@LajuaneTorrey @EytanStarkman @Starbucks @Starbucks_cr I like your phrase, I use a similar one "if you put AI on top of inefficient processes, you'll make inefficiency appear faster"...

@EytanStarkman @Starbucks @Starbucks_cr @RicardoIQSource Exactly this. Every time we've seen large-scale AI fail in retail or ops, the tooling was fine — it was process clarity that wasn't there. You can't automate chaos. You can only expose it faster.

The Pizza Hut + Starbucks pairing is the right comparison. Both deployments failed on the same axis: vendor accuracy claims that were never pressure tested against the actual operating floor before scaling. Combine that with the MIT NANDA finding that 95% of GenAI pilots delivered no measurable impact and you get the structural pattern. Wrote it up here: iqsource.ai/en/blog/starbu…

Starbucks retired an AI inventory tool after frequent mistakes, while Pizza Hut's delivery system allegedly lost a franchisee more than $100 million in sales. f-st.co/dg6jvbB

The real signal here isn't "Starbucks gave up on AI." Niccol is simultaneously expanding Green Dot Assist on Azure OpenAI from a 35-store pilot. The signal is which AI deployments survive contact with the floor, and which ones don't. Acceptable failure mode of the agent matters more than the headline accuracy claim. Full read: iqsource.ai/en/blog/starbu…

JUST IN: Starbucks retires AI inventory tool across North America after it reportedly miscounted & mislabeled store items.