Hello, fellow travelers in the machine learning multiverse.

This week gave us the perfect metaphor for where AI actually stands in late 2025: while Tesla finally secured its robotaxi permit in Arizona—clearing the last regulatory hurdle to launch autonomous vehicles—a tractor company in Idaho is suing Monarch Tractor because their “autonomous” tractors apparently couldn’t autonomously remember how to be autonomous.

The gap between press release and pavement has never been wider.

Between November 18-19, we saw billion-dollar infrastructure deals, million-token context windows, and a16z betting $21M that the most exciting AI application is… tax compliance. (Turns out death and taxes remain undefeated, even by artificial intelligence.)

Let’s dig into the three stories that matter most.

Monarch Tractor Sued Over “Unable to Operate Autonomously” An Idaho dealership is taking Monarch to court, claiming their $100K+ autonomous tractors were defective and couldn’t deliver on the self-driving promise. The lawsuit alleges “significant problems” with all 10 purchased units.

Tesla Receives Arizona Robotaxi Permit Meanwhile, Tesla cleared its final regulatory hurdle to launch robotaxi service in Arizona, a state that’s become the unofficial proving ground for autonomous vehicle dreams.

Why it matters: We’re witnessing the chasm between AI’s potential and its present reliability. While regulators greenlight ambitious deployments, the courtroom is where failed promises go to die. One company’s “breakthrough” is another’s breach of contract.

Read more about Monarch’s lawsuit →

Clem Delangue dropped a truth bomb this week, arguing that all the attention (and capital) is flooding into large language models while smaller, specialized models will ultimately dominate real-world use cases. It’s a classic case of investors fighting over the spotlight while the real work happens in the shadows.

Why it matters: If Delangue is right, a lot of startups burning cash on massive models are in for a rude awakening. The question isn’t whether AI will transform industries—it’s whether the current LLM gold rush is a bubble within a broader revolution.

Read the full interview →

While some warn of bubbles, AI data center provider Lambda just raised a whopping $1.5 billion after landing a multibillion-dollar Microsoft deal. The infrastructure layer is capturing massive value from the AI boom, suggesting that selling pickaxes during a gold rush remains the safest bet.

Why it matters: Someone is clearly betting the bubble has room to inflate. When infrastructure players raise at these valuations, they’re not betting on today’s models—they’re betting on decades of compute demand.

Read more on Lambda’s raise →

Here’s what keeps me up at night: if we’re in an LLM bubble, why are smart money VCs still writing nine-figure checks to infrastructure plays? And why is a16z leading a $21M round for tax compliance software?

Premium subscribers get our full analysis:

[Upgrade to premium for the full breakdown →]

That’s it for the free edition. Want to go deeper? The premium version includes our exclusive “Signal” prediction, a technical deep-dive on why Focal Loss matters for real-world ML, and a curated Q&A about managing AI agents in enterprise settings.

P.S. If you found this valuable, share it with a friend who needs to know what’s actually happening in AI.

Thanks for reading Byte Of Truth! This post is public so feel free to share it.