Number Go Down (NSFW)

Is that the smell of panic in the air?

With the US government no longer a reliable source of employment or economic data, private companies are filling the void:

• ADP reported that 32,000 jobs were lost in November
• Hiring for December also fell short of expectations, even during the holiday season
• Last week, ADP reported hiring in January fell far short of expectations
• In January, layoffs hit the highest level since the Great Recession of 2008-2009

It gets worse:

• A bipartisan group of former agriculture officials warned last week of the potential collapse of American farming due to tariffs and other economic conditions
• The Lever reports that Trump is reverse-engineering the Great Recession of 2008-2009 with a resurgence of mortgage-backed securities and subprime mortgages, calling 2026 “2008 2.0”
• Cryptocurrencies have seen a $2 trillion collapse with many speculating the fault is structural and not cyclical
• In scientific and scholarly publishing, we’re hearing of salaries being frozen, layoffs occurring quietly, and T&E budgets being cut

But wasn’t all this AI stuff supposed to make us rich, the ultimate “deploy and enjoy” scheme?

• Or maybe all this supercharged excitement over LLMs was just a hype bubble or financial black hole created to siphon money from consumer investors and retirement accounts (real dollars) to their own on-paper funny money and stock options?

Techies and their enablers, while lacking true domain expertise, told us that AI could do science, legal, coding, and more. It was futile to call their claims implausible, yet anyone who actually has specialized skills in specialist trades and specialist communities with special certifications and requirements cried foul or knew better. A lot of people with more money than sense fell for it anyhow.

Even the AI homies at the tech companies are beginning to realize they probably fooled themselves into thinking LLMs could generate reliable code. After more than 150,000 layoffs from Big Tech firms based on these promises, and another 30,000 at Amazon last month, it looks like these false claims are going to bite these firms in the butt — 95% of generative AI pilots have failed to deliver measurable ROI despite $40 billion of investment; technical debt is growing due to AI-generated code which tends to be simpler, more repetitive, and less structurally sound; and, the AI slop layer in code repositories is leading to what one developer calls an “AI onslop.”

Even Anthropic developers are starting to break ranks, posting a preprint on arXiv where they discuss how relying on LLMs for coding doesn’t make the work go faster, inhibits learning, and is especially risky in safety-critical domains.

Other concerns center around how entry-level positions are being cut more frequently, creating a potential expertise problem that will manifest in 5-10 years as skills among workers coming out of school simply never have a chance to grow amidst peers.

• One has to wonder if LLMs and AI are so great and powerful, why sell them by subscription? Why not keep them in-house as trade secrets and make money off all the amazing products, businesses, science, efficiency, and patents you’d suddenly be generating?
  • Maybe it’s because these tools aren’t all they’re cracked up to be.

When you peel back the hype and follow the money, lo and behold, the circular financing propping up the larger AI economy becomes apparent — and it is starting to fracture, the sweet aroma of horseshit gradually being supplanted by a growing smell of sweaty panic in the markets.

• Stocks for publicly traded commercial scholarly and scientific publishers are already falling at an accelerating rate, after dropping throughout 2025. Executives and governance bodies seem to be flat-footed and unsure.
  • The days of playing with AI fire seem to be ending with a lot of people about to get burned.

For the larger economy, tech stocks are starting to crumble as the promise of AI is coming up snake eyes: