惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

B
Blog RSS Feed
有赞技术团队
有赞技术团队
OSCHINA 社区最新新闻
OSCHINA 社区最新新闻
阮一峰的网络日志
阮一峰的网络日志
V
V2EX
Y
Y Combinator Blog
Jina AI
Jina AI
G
Google Developers Blog
Last Week in AI
Last Week in AI
博客园 - 叶小钗
H
Hackread – Cybersecurity News, Data Breaches, AI and More
L
LangChain Blog
奇客Solidot–传递最新科技情报
奇客Solidot–传递最新科技情报
V
Visual Studio Blog
aimingoo的专栏
aimingoo的专栏
freeCodeCamp Programming Tutorials: Python, JavaScript, Git & More
J
Java Code Geeks
Stack Overflow Blog
Stack Overflow Blog
IT之家
IT之家
The GitHub Blog
The GitHub Blog
D
Docker
量子位
罗磊的独立博客
腾讯CDC

The Decoder

The AI industry's platform trap is starting to look a lot like Microsoft's OpenAI buys Ona to push Codex toward long-running, autonomous coding tasks Jeff Bezos' AI startup Prometheus closes $12 billion round at a $41 billion valuation Free Deezer tool lets users on any streaming service check their playlists for AI music OpenAI vs. Anthropic: A price war over API tokens is brewing Dario Amodei's new essay reads like a Cold War playbook for the AI age Claude Fable 5: Anthropic admits "wrong tradeoff" after invisibly throttling rival AI researchers Google's new open model DiffusionGemma generates text from noise instead of word by word OpenAI's IPO slips as Altman tells staff to expect a public offering "within the next year" Anthropic study shows AI needs hours, not weeks, to build exploits from security patches OpenAI wants its biggest data center yet, and Nvidia would back the bill Claude Fable 5: The first Mythos model is powerful, expensive, and heavily filtered Germany's National Security Council greenights an AI Safety Institute modeled after the UK's AISI Google's NotebookLM now runs its own cloud computer with code execution and agent-based research Anthropic releases Claude Fable 5 and Mythos 5 with major gains in coding and science Google's Gemini 3.5 Live Translate delivers real-time voice translation across 70+ languages SpaceX wants to put data centers in orbit, and Musk says it's no big deal Landmark German ruling declares Google's AI Overviews are Google's own words and makes it liable for false answers Beijing's $295 billion AI buildout would require 80 percent domestic chips, locking out US suppliers Apple Intelligence gets a second shot with help from Google and Nvidia OpenAI now says "entirely automating everything is not the future we want" OpenAI says going public is "a complicated set of tradeoffs" and is unsure about the timing Microsoft Research's Lens proves detailed captions matter more than raw scale for training efficient image generators Intel gets a second life as Google and Nvidia explore it as a TSMC backup for AI chips Most companies are flying blind on AI spending Frontier Radar #3: How agentic AI is turning tokens into a business metric Instagram AI chatbot breach may have affected over to 20,000 accounts, Meta discloses Microsoft tightens rules for conflict zones after investigation into Israel's military use of Azure Moonshot AI targets a $30 billion valuation, more than six times its late-2025 worth Deepseek topped Ramp's trending software vendors in June 2026 as US companies chase cheaper AI
Meta shifts from "tokenmaxxing" to token managing as inte...
Matthias Bastian · 2026-06-13 · via The Decoder

Meta is allegedly running up billions in AI spending and now wants tighter oversight of how AI gets used, by whom, and to what end.

In an internal memo sent to about 6,000 employees, Meta flagged an "exponential increase" in AI usage and warned the company is on track for billions in costs from internal use alone by 2026, The Information reports. Individual employees and teams had no visibility into, or control over, their own consumption.

Starting in 2027, Meta plans to manage AI tokens more tightly with budgets, allocations, and dedicated tools. A team of developers and engineers built a central dashboard called "AI Gateway" that tracks usage and spending in one place.

Automatic alerts for unusual cost spikes are coming next. Meta also wants to steer employees away from third-party tools like Anthropic's Claude and toward its own coding assistant, MetaCode. Other models will still be available, though; Meta's own models aren't yet competitive at the frontier.

Token usage doesn't equal productivity

Engineers in Meta's new "Applied AI Engineering" division are working to improve MetaCode by creating coding tasks as training data. Earlier, Meta had made AI usage a "core expectation" in performance reviews, which led to so-called "tokenmaxxing": employees artificially inflated their consumption through an internal leaderboard called "Claudeonomics," racking up 73.7 trillion tokens in just over 30 days.

CTO Andrew Bosworth pushed back in a separate memo: "Nobody should be using AI tools just for the sake of using them. All motion is not progress and token usage alone is not a measure of impact of any kind." Tools should be used when they "genuinely allow us to do better work, faster."

Amazon ran into a similar tokenmaxxing problem that spiraled out of control. That both companies are now reining in AI spending fits a broader pattern: businesses are questioning whether AI is actually boosting productivity. Sam Altman recently called cost control a "huge issue" among his customers, likely driven in part by massive price hikes for model usage.

AI News Without the Hype – Curated by Humans

Subscribe to THE DECODER for ad-free reading, a weekly AI newsletter, our exclusive "AI Radar" frontier report six times a year, full archive access, and access to our comment section.

Subscribe now