PierG

@burkov next time we wait for you in Italy: we always find a way …. 😎😂

Just arrived in Lyon at a 4-star hotel with "accessible" parking advertised online. The parking door is closed, and no one answers the intercom for 15 minutes. The street is no-parking. In Lyon, parking in the city is crazily complex and expensive. The price depends on the zone and the type of vehicle. You don't know what type your vehicle is, and the city's website doesn't define it clearly. So, we assumed that the highest rate of 80 Euros per 10 hours of parking applies (as indicated on the parking meter). Leaving a car far from the hotel and paying 80 Euros per 10 hours sounded crazy. After 15 minutes of calling on the hotel's intercom, the receptionist finally answered, said, "If I haven't answered, that means there are no places," and hung up. I called again and asked, "Where do I put my car then?" He answered, "Wherever you want," and hung up again. Lyon, oh mon amour!♥️

@joshwoodward What about Enterprise users?

We wanted to get this to you ASAP, and there are a few known issues: * Veo 3 in the Gemini Pro plan works on the web, with mobile and Workspace coming soon * We officially support English-only audio output, but you might occasionally hear another language * Gemini web users on Pro can switch to Veo 2 on mobile if your Veo 3 credits run out * Flow currently doesn’t support voice output for image uploads Let us know if you find other rough edges, and please keep generating responsibly! Enjoy!

Veo 3 dropped about 100 hours ago, and it's been on 🔥🔥🔥 ever since Now, we’re excited to announce: + 71 new countries have access + Pro subscribers get a trial pack of Veo 3 on the web (mobile soon) + Ultra subscribers get the highest # of Veo 3 gens w/ refreshes How to try it… ➡️ Gemini (gemini.google): * Great for everyone - click the Video chip in the prompt bar, and just describe your video * Pro subscribers now get a 10-pack so you can try it * Ultra: MAX limits, daily refresh! ➡️ Flow (flow.google): * Great for AI filmmakers * Pro: 10 gens/month * Ultra: Now 125 gens/month (up from 83)!

@joshwoodward @GeminiApp in Gems execute the instructions when you create a new chat (like for example for introductory instructions or something)

✅ Papercut fixed: @GeminiApp now lets you upload multiple files at the same time. Sorry we didn't have this before - if you see other little annoying things like this, drop them here! We'll aim to fix the most common things that cause grumpiness.

@burkov @pvncher coupling / cohesion at a different level (kind of)

It's a nice trick, but it's just a band-aid. Eventually, even the relevant pieces will reach 100k. What we need, in the absence of effective 1M or 10M context support, is a new way of engineering apps, where most components are tiny and entirely unrelated to one another. Similar to microservices but for the UI-heavy apps.

100k tokens. This is where the excitement from LLM-produced code turns into frustration, no matter the LLM. Grok just doesn't see anything beyond that threshold. Claude just provides guesses. Gemini rewrites everything every time, and it just gets uglier and doesn't work. No serious application's codebase is under 100k tokens long. I doubt that any substantial LLM improvement will allow it to be effective with a 1M token codebase. Until then, vibe coding is for prototypes and proofs of concept only. Devin, you are our last hope! Don't disappoint us.

@GeminiApp as Enterprise Workspace subscribers we are still waiting...

The roll-out for video generation (Veo 2) is now complete. Type a prompt—or even just an emoji—and get a video in minutes. Here's how: 1️⃣ Download the Gemini app or head to gemini.google.com 2️⃣ Sign up for Gemini Advanced (eligible students get 12 months FREE! 🎓) 3️⃣ Select 'Veo 2' from the model drop-down Prompt for this video: 📦🐸🍄 as mascots More examples in 🧵

@vincos @e_pagliarini yes... oltretutto caso scuola in tanti corsi di astuzia e furbizia di innovazione.

@e_pagliarini mizzica, sembrava un ambito promettente

Boom!

@Ferdinando1970 Kind of... "È un test, a un operatore non sofisticato direi: non lo fate mai, non lo faccio io, non lo farei fare a mia madre"

C’è Intesa su Bitcoin, si è convertita come Sanpaolo sulla via di Washington

@CharlesOnwuakpa 😰

No, non sto piangendo, è che mi è entrato un razzo nell'occhio 🚀😅.

"Sono riusciti a re-infilare un ago nell'asola dopo che ha fatto un giretto nell'universo" (cit. Francesca Mingardi) #SpaceX

Vi aspettiamo a Modena 😊 !!! @credem @CredemtelEventi @PierG #futuro #innovazione #wellbanking

@raiza_abubakar Really interesting feature Raiza: congrats. It might be interesting to be able to give a sort of "system prompt" to the system in general (not just for the podcast part)

Kinda crazy last 24h with this “jailbreak” using sources to get this kind of output. Here are my reactions: 1) AOs are designed to give you magic in exchange for a little bit of content. Give it a source and let it do the rest for you — it’s kind of unpredictable (and seemingly delightful) based on the stuff I’ve seen people share (upload resume = get a hype crew). We’re going to lean in here and try to accelerate more of this type of stuff in the product. 2) After that initial moment of delight, people want to influence the content. Makes sense - now you want knobs for format, length, personas, voices, languages - I’m keeping track and we’re working on it (some will come much faster than others). 3) The content studio we’ve built + the voice model makes really powerful, emotive stuff. You can see it can take on a really upbeat hyper-interested tone (literally for any topic), but now we’ve seen the flip side. I was literally 😭 when Sam said he was going to call his wife! Need to think about this more. 4) The world has moved so quickly (and our shared internet consciousness?) just from what I’ve seen here and on Reddit. There’s far fewer extreme reactions and more of the “oh how’d you do that” and “it’s just creating based on a source.” 5) It’s been really fun listening to what everyone’s been making :) Thanks for sharing all your AOs. I can’t believe this is what got me to come back on Twitter / X lol

Vera Gheno a #OfficineCredem mi fa ricordare quanto fosse bello il Twitter... @credem

@tg2rai era davvero fondamentale intervistare il 18 enne che guidava il trattore?

Life at #OfficineCredem

@gicappa ... e un filino complesso. L'algoritmo di detection della velocità di lettura del Kindle mi dice che a finirlo ci metterò infiniti giorni 😃

@PierG L’ho letto un po’ di anni fa ormai ma rimane sempre un testo illuminante.

… e qui iniziò il casino. leggi.amazon.it/kp/kshare?asin…

Questa cosa che non facciano scarpe da montagna / trail senza supporto per gli iperpronatori è inaccettabile... #trail

“For free…” uuuuu #WWDC24