What I Learned About AI Trust from Reconciling over 100B Transactions • Hacker News Board

Re: What I Learned About AI Trust from Reconciling over 100B Transactions

#21

Post by pavel_lishin » Mon, Aug 31, 2026, 2:05 PM UTC

joebates wrote: post #16

Insensitivity wrote: post #2

The smooth scroll / animation or whatever it is makes it so hard to read / use

I take a small bit of joy in disabling JS for these sites. The content still loads just fine. The scroll jacking degraded quality for no reason. Why would anyone who actually wants me to read their site do this?

They didn't, they asked an LLM to create a webpage, and it did, and they didn't notice that particular bit of sloppery, any more than people who post AI-generated memes on Facebook don't notice that their niece has six fingers now.

Re: What I Learned About AI Trust from Reconciling over 100B Transactions

#22

Post by oefrha » Mon, Aug 31, 2026, 2:20 PM UTC

This reads like someone took two conventional non-AI reports about (1) different teams’ conflicting viewpoints of the same data; (2) deterministic traceability that is hardly remarkable, and asked Claude to merge the two reports and shoehorn some disjoint AI narratives into it. And did a very bad job of it. What the hell is this incoherent nonsense? AI writing aggregate SQL queries (“what is this month’s revenue by entity?”) is completely reliable because every tx is traceable, but wouldn’t be reliable if every tx is not traceable, seriously?

Re: What I Learned About AI Trust from Reconciling over 100B Transactions

#23

Post by huurtehoog » Tue, Sep 01, 2026, 2:29 AM UTC

giancarlostoro wrote: post #14

vintermann wrote: post #7

Earlier quoted context omitted.

"What I Learned About UI Design from Getting Roasted on Hackernews"

Sadly they never do, UI just gets worse and worse. Even before LLMs became as prevalent, and even after. ;)

Good code enables easy code. Easy code engenders bad code. Bad code necessitates good code.

For all the 'revolutions' in programming computers, DARPA is hailing formal methods as the big thing in computer security now.

Boom and bust, as in any distributed system with signalling delays. "Oh boy here I go mathin' again" =]