Gioele Zorzi

Is Threads the official headquarters of anti-AI, anti-progress, technophobic cavemen, or am I just unlucky with the algorithm?

Every generation has its version of "this technology will ruin us." The steam engine. Photography. The telephone. The internet. Now AI. They were wrong every time, just saying.

@elkelk "here we like to do things differently"

Google shoving a square UI onto a circular watch

@KKengo2 This is so accurate

Gemini 3.1 Pro is completely broken since the platform update. We’re paying more for tighter compute limits, only for the model to hallucinate wildly and completely fail to read basic chat context. I don't know what they did to the backend, but they massacred my boy.

@wingedwork Some serious obfuscated enshittification happing there

@twentythrill Google has done something that could be considered one of the most outrageous things in AI history: they significantly degraded a model's performance to roughly 30% of its original level without even changing the version number.

@MartinMLynch I'm currently testing Xiaomi MiMo 2.5 Pro and I'm quite happy with it, it's way cheaper and on paper just a touch less "smart" than Gemini 3.1 Pro

@twentythrill Your description matches our experience. We're hitting rate limits inside 1 hour of coding while having to routinely generate 4-7 iterations to correct all the breaking mistakes Gemini 3 series introduces. It's terrible. Reluctantly exploring alternatives. Any suggestions?

@oh_meed I tried it a lot, but definetly not on par intelligence, hallucinating a lot 😭

@twentythrill Ok but why not use 3.5 flash? It’s faster, on par intelligence wise and in the same price class.

@D0miniqZ Yep for me also, the more verbose, the more it tent to hallucinate

@twentythrill In my expierience, 3.5 flash high is the one that’s hallucinating the most. 3.5 med doesn’t hallucinate as much. 3.1 Pro on both settings is the most stable out of all, I’d say (well still not as stable as Opus 4.6).

@Maz628037437162 Also I confirm 3.5 Flash hallucinate a lot on programming troubleshooting

@twentythrill So far, I’ve found the 3.1 Pro version better than the 3.5 Flash. The Pro fixed an issue with my HTML file while the 3.5 Flash completely redesigned the page and broke all features even after I requested specific changes and instructed it to avoid affecting anything else.

Dear @Google, how is a monthly spend cap for a paid production API labeled as an "Experimental" feature? 🫠

Yes, Liquid Glass still sucks, even on macOS. Latest Tahoe 26.4.1 - not a beta! :D

Another bait-and-switch: Gemini changed their paid plans to "compute-based" limits. Three questions in and I hit 34% of my daily cap, plus 8% of a weekly limit. Basically, a limit on top of a limit! Subscription canceled soon🖕 #GoogleIO2026 #GoogleIO #Gemini

I want to like Liquid Glass, but Apple missed the mark. Nearly 10 months in, and it's still a laggy, inconsistent mess. It’s the perfect example of a cool visual concept that completely falls apart under real-world performance and legibility requirements.

At first, I thought this was a joke. But apparently it's on SONY official page...

@ShishirShelke1 Apple really forgot what good UI design is. RIP.

This is what iOS Control Centre looked like just 2 years ago… Crazy how outdated it feels now 😭

BTW OpenAI dropped GPT 5.5 but is still one of the worst at avoiding hallucinations 🤦🏻 Source: artificialanalysis.ai

Hardware should be fun to use. ​Developed an RSVP reading app that finally makes use of the R1’s scroll wheel. Scrolling words back and forth feels incredibly natural. Why isn't every "reading" experience this tactile?

I'm pathologically incapable of doing things manually Just spent 30 mins vibe coding an outbound system so I never have to send a manual email again. #VibeCoding #BuildInPublic #AI #Automation #Solopreneur #Outbound