Bert Hubert is one of those rare technologists whose writing consistently rewards the time spent reading it. He's the author of PowerDNS, a former member of the Dutch intelligence oversight committee, and a prolific essayist on the seams where computing, biology, and policy meet. When he publishes something with a title like "The Impossible Things We Have to Believe," it's worth pausing for.

The phrase itself echoes the White Queen in Through the Looking-Glass, who practiced believing six impossible things before breakfast. Hubert's essays in this vein typically catalogue the load-bearing fictions modern technical society depends on — claims that fall apart under scrutiny but which we accept because the alternative is paralysis. Past pieces of his have dissected things like:

• The notion that complex software supply chains are auditable in any meaningful sense
• The fiction that "the cloud" is a stable substrate rather than a leased dependency
• The claim that cryptographic guarantees survive contact with real-world key management
• The assumption that biological systems are well-enough understood to engineer reliably

Given the timing, this post almost certainly engages with the current AI moment — the gap between what large models are claimed to do and what they actually do, the impossible-to-verify training data provenance, the assertion that scaling alone will produce reasoning. But Hubert rarely writes pure polemic. His angle tends to be: here are the specific places the abstraction leaks, and here is what an honest engineer should do about it.

Why this deserves a technical audience's attention: most discourse about overpromising in tech is either credulous boosterism or reflexive cynicism. Hubert occupies a much rarer position — he builds infrastructure that real systems depend on, he's seen government and standards processes from the inside, and he writes with the precision of someone who knows that vague critiques are useless. His essays often surface specific, falsifiable claims you can take back to your own work and ask: am I assuming this too?

The piece is also from a personal blog, which on HN means it competes against polished corporate content marketing for attention and usually loses — exactly the kind of long-tail submission that the front-page algorithm tends to bury.