From Tokenmaxxing to Tokenminning. Coinbase’s CEO just published a post about how they cut their AI spend in half by defaulting to open weight Chinese models like GLM 5.2 and Kimi 2.7.
As AI vendors stop subsidizing tokens, setups like this will become common as AI becomes another cost to manage.
As AI vendors stop subsidizing tokens, setups like this will become common as AI becomes another cost to manage.
June 27, 2026 at 2:50 PM
From Tokenmaxxing to Tokenminning. Coinbase’s CEO just published a post about how they cut their AI spend in half by defaulting to open weight Chinese models like GLM 5.2 and Kimi 2.7.
As AI vendors stop subsidizing tokens, setups like this will become common as AI becomes another cost to manage.
As AI vendors stop subsidizing tokens, setups like this will become common as AI becomes another cost to manage.
Going from tokenmaxxing to tokenminning is no longer just for developers. Microsoft is allegedly updating the AI models used by Excel and Outlook from OpenAI & Anthropic's models to their homegrown AI models to reduce costs.
I suspect this is likely more about dogfooding internal models than costs.
I suspect this is likely more about dogfooding internal models than costs.
July 7, 2026 at 6:00 PM
Going from tokenmaxxing to tokenminning is no longer just for developers. Microsoft is allegedly updating the AI models used by Excel and Outlook from OpenAI & Anthropic's models to their homegrown AI models to reduce costs.
I suspect this is likely more about dogfooding internal models than costs.
I suspect this is likely more about dogfooding internal models than costs.
i keep seeing the word "tokenminning" thrown around the work group chats and need to do a murder
August 4, 2026 at 9:57 PM
i keep seeing the word "tokenminning" thrown around the work group chats and need to do a murder
Tokenminning
“Companies across tech, entertainment, banking, and many other industries are throttling their employees’ use of AI and pleading with workers to use less powerful models to stop AI costs from spiraling out of control…”
Companies Are Throttling Employees’ AI Use Because It’s Too Expensive
Sources and leaks from Amazon, Adobe, Atlassian, Citi, and more show what is really happening with AI right now: companies are trying to reign in AI use as costs spiral out of control.
www.404media.co
July 2, 2026 at 11:29 AM
Tokenminning
Sticker shock has execs rethinking this whole AI thing
Sticker shock has execs rethinking this whole AI thing
This week on The Reg's Kettle podcast, we wonder whether tokenminning is going to bring the industry back down to Earth
www.theregister.com
July 15, 2026 at 10:19 PM
Sticker shock has execs rethinking this whole AI thing
TOKENMINNING....is that anything
May 19, 2026 at 5:44 PM
TOKENMINNING....is that anything
More details on the death of tokenmaxxing and the rise of tokenminning below
www.wsj.com/business/chi...
www.wsj.com/business/chi...
Corporate America Has Suddenly Decided to Stop Blowing Money on AI
Companies big and small are mixing models, and it’s changing the economics and power players of the industry.
www.wsj.com
July 26, 2026 at 1:08 AM
More details on the death of tokenmaxxing and the rise of tokenminning below
www.wsj.com/business/chi...
www.wsj.com/business/chi...
Now: Tokenminning, Meta, Uber, Walmart and Amazon edition.
www.nytimes.com/2026/06/18/t...
www.nytimes.com/2026/06/18/t...
June 24, 2026 at 8:14 PM
Now: Tokenminning, Meta, Uber, Walmart and Amazon edition.
www.nytimes.com/2026/06/18/t...
www.nytimes.com/2026/06/18/t...
Sticker shock has execs rethinking this whole AI thing
Sticker shock has execs rethinking this whole AI thing
This week on The Reg's Kettle podcast, we wonder whether tokenminning is going to bring the industry back down to Earth
www.theregister.com
July 14, 2026 at 2:18 PM
Sticker shock has execs rethinking this whole AI thing
Sticker shock has execs rethinking this whole AI thing
Sticker shock has execs rethinking this whole AI thing
This week on The Reg's Kettle podcast, we wonder whether tokenminning is going to bring the industry back down to Earth
www.theregister.com
July 13, 2026 at 2:18 PM
Sticker shock has execs rethinking this whole AI thing
Tokenmaxxing was trying to use AI as much as you can, since everyone had large fixed monthly quotas
Now that prices have greatly increased, Tokenminning is trying to get as much value from AI while costing as little as possible
It’s the latest thing in Silicon Valley*
Now that prices have greatly increased, Tokenminning is trying to get as much value from AI while costing as little as possible
It’s the latest thing in Silicon Valley*
June 10, 2026 at 10:45 PM
Tokenmaxxing was trying to use AI as much as you can, since everyone had large fixed monthly quotas
Now that prices have greatly increased, Tokenminning is trying to get as much value from AI while costing as little as possible
It’s the latest thing in Silicon Valley*
Now that prices have greatly increased, Tokenminning is trying to get as much value from AI while costing as little as possible
It’s the latest thing in Silicon Valley*
‘Earlier this year, the message from tech companies to employees was clear: Use as much artificial intelligence in your work as possible. Now the tokenmaxxing era appears to be over. ”Tokenminning,” short for “token minimizing,” is now in.‘ www.nytimes.com/2026/06/18/t...
Tech Workers Maxed Out Their A.I. Use. Now They’re Trying to Minimize It.
www.nytimes.com
June 19, 2026 at 4:37 PM
‘Earlier this year, the message from tech companies to employees was clear: Use as much artificial intelligence in your work as possible. Now the tokenmaxxing era appears to be over. ”Tokenminning,” short for “token minimizing,” is now in.‘ www.nytimes.com/2026/06/18/t...
'The desire to acquire your most valuable knowledge has only increased as the agentic AI revolution fails to take off. The recent price rises make AI as expensive as humans. Suddenly, “tokenmaxxing” is out, and “tokenminning”, which means using AI as sparingly as possible, is all the rage.'
June 24, 2026 at 8:07 AM
'The desire to acquire your most valuable knowledge has only increased as the agentic AI revolution fails to take off. The recent price rises make AI as expensive as humans. Suddenly, “tokenmaxxing” is out, and “tokenminning”, which means using AI as sparingly as possible, is all the rage.'
Sticker shock has execs rethinking this whole AI thing www.theregister.com/ai-and-ml/20...
Sticker shock has execs rethinking this whole AI thing
This week on The Reg's Kettle podcast, we wonder whether tokenminning is going to bring the industry back down to Earth
www.theregister.com
July 13, 2026 at 5:50 PM
Sticker shock has execs rethinking this whole AI thing www.theregister.com/ai-and-ml/20...
Your AI bill has a hidden lever: quantization.
Model weights carry more precision than needed. Round them down (FP16→INT8→INT4) and one benchmark saw a 70B model's GPU cost drop from $24K to $4K. Same model, 83% cheaper.
sgopalanbtg.substack.com/p/explainer-...
#TokenMinning #AI #FinOps
Model weights carry more precision than needed. Round them down (FP16→INT8→INT4) and one benchmark saw a 70B model's GPU cost drop from $24K to $4K. Same model, 83% cheaper.
sgopalanbtg.substack.com/p/explainer-...
#TokenMinning #AI #FinOps
Explainer: Quantization - The Rounding Trick That Cuts Your AI Bill by 75%
Sriram Gopalan's thoughts on AI
sgopalanbtg.substack.com
September 21, 2026 at 2:28 PM
Your AI bill has a hidden lever: quantization.
Model weights carry more precision than needed. Round them down (FP16→INT8→INT4) and one benchmark saw a 70B model's GPU cost drop from $24K to $4K. Same model, 83% cheaper.
sgopalanbtg.substack.com/p/explainer-...
#TokenMinning #AI #FinOps
Model weights carry more precision than needed. Round them down (FP16→INT8→INT4) and one benchmark saw a 70B model's GPU cost drop from $24K to $4K. Same model, 83% cheaper.
sgopalanbtg.substack.com/p/explainer-...
#TokenMinning #AI #FinOps
There's a lot of misguided focus on AI 'tokenmaxxing' – but how good are you at 'tokenminning'? New post: kucharski.substack.com/p/how-good-a...
How good at you at ‘tokenminning’?
Have a go at tightening your AI belt
kucharski.substack.com
May 16, 2026 at 12:18 PM
There's a lot of misguided focus on AI 'tokenmaxxing' – but how good are you at 'tokenminning'? New post: kucharski.substack.com/p/how-good-a...
De AI‑hausse kantelt. Big Tech draait de kraan dicht: kosten exploderen, winst blijft achter. Van tokenmaxxing naar tokenminning — niet omdat het kan, maar omdat het móét.
We zitten midden in een reality check: #AI is geen gratis wondermachine, maar een dure afhankelijkheid.
We zitten midden in een reality check: #AI is geen gratis wondermachine, maar een dure afhankelijkheid.
Tech Workers Maxed Out Their A.I. Use. Now They’re Trying to Minimize It.
www.nytimes.com
June 18, 2026 at 3:00 PM
De AI‑hausse kantelt. Big Tech draait de kraan dicht: kosten exploderen, winst blijft achter. Van tokenmaxxing naar tokenminning — niet omdat het kan, maar omdat het móét.
We zitten midden in een reality check: #AI is geen gratis wondermachine, maar een dure afhankelijkheid.
We zitten midden in een reality check: #AI is geen gratis wondermachine, maar een dure afhankelijkheid.
"Tokenminning is a new pattern, which systematically minimizes token use while maintaining, if not improving, the performance of your AI agents."
Sam Black explains why we should move beyond the excesses of tokenmaxxing and think more explicitly about efficiency.
Sam Black explains why we should move beyond the excesses of tokenmaxxing and think more explicitly about efficiency.
Tokenminning: How to Get More from Your Chatbot for Less | Towards Data Science
Tokenmaxxing is out. Real patterns for reducing costs without sacrificing AI effectiveness
towardsdatascience.com
August 11, 2026 at 12:34 AM
"Tokenminning is a new pattern, which systematically minimizes token use while maintaining, if not improving, the performance of your AI agents."
Sam Black explains why we should move beyond the excesses of tokenmaxxing and think more explicitly about efficiency.
Sam Black explains why we should move beyond the excesses of tokenmaxxing and think more explicitly about efficiency.
Your AI's "on second thought" isn't reconsideration. It's autocomplete predicting a correction pattern learned from humans thinking aloud. No draft, no delete key — you pay for the wrong turn AND fix.
sgopalanbtg.substack.com/p/the-actual...
#AI #FinOps #TokenMinning
sgopalanbtg.substack.com/p/the-actual...
#AI #FinOps #TokenMinning
The 'Actually, On Second Thought' Tax — And How to Stop Paying It
Sriram Gopalan's thoughts on AI
sgopalanbtg.substack.com
September 19, 2026 at 4:36 PM
Your AI's "on second thought" isn't reconsideration. It's autocomplete predicting a correction pattern learned from humans thinking aloud. No draft, no delete key — you pay for the wrong turn AND fix.
sgopalanbtg.substack.com/p/the-actual...
#AI #FinOps #TokenMinning
sgopalanbtg.substack.com/p/the-actual...
#AI #FinOps #TokenMinning
Tokenmaxxing > Tokenminning
Use Haiku for lookups, rewrites, single tool calls, summaries.
Use Sonnet for reasoning, code-gen, and analytics
Use Opus for multi-agent and outcome-based.
Use Haiku for lookups, rewrites, single tool calls, summaries.
Use Sonnet for reasoning, code-gen, and analytics
Use Opus for multi-agent and outcome-based.
June 18, 2026 at 8:00 AM
Tokenmaxxing > Tokenminning
Use Haiku for lookups, rewrites, single tool calls, summaries.
Use Sonnet for reasoning, code-gen, and analytics
Use Opus for multi-agent and outcome-based.
Use Haiku for lookups, rewrites, single tool calls, summaries.
Use Sonnet for reasoning, code-gen, and analytics
Use Opus for multi-agent and outcome-based.
And that is the limit of how much recap I am inspired to compile. Also I need to go to a work where we were just told, in short order, "do tokenmaxxing", "no wait not like that", "ok, now do tokenminning and produce the same results".
August 20, 2026 at 2:10 PM
And that is the limit of how much recap I am inspired to compile. Also I need to go to a work where we were just told, in short order, "do tokenmaxxing", "no wait not like that", "ok, now do tokenminning and produce the same results".
Tell them about “Tokenminning” - my original title for this article.
www.linkedin.com/posts/avik-d...
www.linkedin.com/posts/avik-d...
Token Engineering: The New Compute Efficiency | Avik Dey
In the cloud era, the winners were the companies that mastered compute efficiency by right sizing instances, autoscaling and spot fleets. In the LLM era, the equivalent discipline is token efficiency....
www.linkedin.com
July 25, 2026 at 3:56 PM
Tell them about “Tokenminning” - my original title for this article.
www.linkedin.com/posts/avik-d...
www.linkedin.com/posts/avik-d...
"The reversal [from tokenmaxxing to tokenminning], within just a few months, underlines how AI use remains in flux as people try to figure out how to best use the tools."
“People” or corporate executives committed to the ideology of AI?
www.nytimes.com/2026/06/18/t...
“People” or corporate executives committed to the ideology of AI?
www.nytimes.com/2026/06/18/t...
Tech Workers Maxed Out Their A.I. Use. Now They’re Trying to Minimize It.
www.nytimes.com
June 19, 2026 at 1:38 PM
"The reversal [from tokenmaxxing to tokenminning], within just a few months, underlines how AI use remains in flux as people try to figure out how to best use the tools."
“People” or corporate executives committed to the ideology of AI?
www.nytimes.com/2026/06/18/t...
“People” or corporate executives committed to the ideology of AI?
www.nytimes.com/2026/06/18/t...