ABZI

@launch_llama abzi.ai - check it out, no pressure!

I want to be someone's first customer. Drop ur startup link

@DanielSmidstrup Trillions of parameters = 💵💵💵 though as they get more efficient in the long run I'm sure it will get better. Or if their users don't pay and they are forced to lower costs. 😉

Weren’t AI prices supposed to trend toward zero? Why does every new model launch with a higher price tag?

@Google I'll give flash lite a try. Looking for a decent haiku replacement. Please keep releasing more affordable models (and stick to Gemma releases for local open source, I think these are more important than anything).

Gemini 3.5 Flash-Lite: Our fastest and most cost-effective model in the 3.5 family. It significantly outperforms 3.1 Flash-Lite across thinking levels, coding, and agentic tasks.

Today we’re expanding the Gemini family with three new models built to be faster, more token efficient, and reliable at scale. Meet the new Gemini models ↓

@LuminaXspace @JAKEVIN16 Hope it's good. 3.1 Pro made a big splash at the time, but in the AI world that accounts for very little these days.

@JAKEVIN16 probably

🚨 Google has officially started training Gemini 4 What is confirmed: • Gemini 4 pre-training is now underway • Google describes it as its most ambitious training run ever • It will form the foundation for Google’s next generation of models • Development is happening alongside Gemini 3.5 Pro partner testing • No release date, benchmarks or technical specifications have been announced This is still only the pre-training stage. After the initial run, Gemini 4 will require extensive post training, reasoning optimisation, safety testing and external evaluations before release. Could Gemini 4 become Google’s real attempt to retake the frontier?

@Da7_Tech What kind of 4B you using?? Lol agreed that it's not great.

I gave Gemini 3.6 Flash a medium-sized task and expected it to spend a reasonable amount of time on it. It finished in two minutes. At first, I thought the speed was insane. Then I saw the result. A local 4B model could do better. Do they have no shame at Google?

@bridgemindai Honestly, the 3.5 flash model was "fine". Fast, overpriced, but gets the job done for normal everyday stuff. Probably should have held off until they had something worth a headline though. I can hear the panicked executive saying, "Just release something!"

Gemini 3.6 Flash just dropped and it is the EXACT same model as 3.5 Flash. 50 on the Intelligence Index. 3.5 Flash: also 50. Identical intelligence, slightly faster. Another flop from Google. They shipped a speed patch and called it a version. Six months of silence for THIS? Nobody wants a faster Flash. Everyone wants Gemini 3.5 Pro. Ship the real one, Google.

Revolutionary new AI infrastructure: the laptop you already own. Try today.

@thearslaniqbal Great, my kids love Frozen 2!

🚨 Google is reportedly building a new AI chip called Frozen v2, expected to launch in 2028. The goal is simple... Make Gemini AI much faster and more power-efficient. Reports say the new chip could be 6–10x more efficient than Google's current AI chips. Big AI companies are now designing their own chips instead of relying only on NVIDIA. OpenAI, Anthropic, and now Google are all racing to build the next generation of AI hardware.

@Akshay272727 Let's connect 👋

Hi, I started with 50 followers yesterday, now at 120🤯. Feels surreal meeting so many awesome folks. But I don't want it to stop... If you're in: - Tech - AI - Data - SaaS - Startups - Development Would love to know what you're working on. Say hi and let's grow together 👋

I have 1,200+ followers. When I started with nothing I looked at accounts with 1,200+ followers and thought man... How did they do it? Those dudes are fucking famous. Now I look at bigger accounts. Dudes with 100k+. And now I think, I wonder when I'll hit 100k. Shit can be done if you just try and believe in yourself.

@Technouniverse9 I don't think they're in a position to get it wrong right now. Better to not release anything until it's ready.

🚨 Gemini 3.5 Pro Update: The July 17 launch that never happened. What we know now: > Google reportedly scrapped the ENTIRE base model and rebuilt from scratch > Bloomberg: delayed again over hallucination rates & reliability, 3rd slip since June > Leaked internal evals still claim it beats GPT-5.6 Sol and Claude Fable 5 > Rumored specs: 2M token context, Deep Think reasoning, major frontend/SVG gains > Internal codename: Cappuccino ☕ > Prediction markets now bet on late July / early August Google announced this thing at I/O in May. Meanwhile GPT-5.6 and Grok 4.5 already shipped. Is Google being perfectionist... or is something wrong? 👀

Hey I'm Chris, Just a dad building abzi.ai in my spare time (which doesn't exist). Looking to connect with people building in: • Local AI • Open Source • Dev Tools • AI Orchestration Reply with what you're working on and let's connect! 👇

@Polymarket Next up: Banning Chinese-made components from the actual GPUs running the AI! 🫠Come on..

JUST IN: Trump officials are reportedly weighing blocking U.S. access to top Chinese AI models.

@Ra3orbladez Let's connect 💪

I need ~300 more followers to 2k, so let's #connect if you're into development, AI, open-source, data security and local-first software.

@shergilldotdev Love Totoro 👋

My original goal on X was to connect and talk to likeminded people. I wanted to learn and grow. If you are apart of the 🧑‍💻 tech space… Let’s connect!

I think it will just be a normal thing like anything else in the future. People only care about the final product and the value it brings them. As long as it's genuine, that will come across. Yes, there are a subset that will gravitate to the hard made stuff, but even that bar has changed over time. Do you care about a book that was typed on a typewriter or computer? The tools are the tools. Quality will determine the rest.

random thought: will “no ai features” become the new premium feature one day? 👀

@jun_song That's exactly why it's not transparent. It's on purpose so they can control whatever they want and change things on the fly.

I think GPT and Claude need to be way more transparent about their subscription usage limits. None of their plans clearly state how many tokens we actually get for what we pay. Calling plans "X5" or "X20" isn't enough when we don't even know what the baseline allocation is in the first place.

@TimJayas People be like.

why do people pay $200 for Fable and GPT-5.6 Sol Ultra only to hit weekly limits in a single day?

Let's not forget that OpenAI started as a non-profit with one goal: benefit humanity. They changed that mission statement six times in nine years. In 2025 they removed the word "safely" entirely. You can decide what that means. Everybody always starts off with the greatest of intentions. It's honestly a lesson for all of us.

Anthropic literally broke off from OpenAI for "AI safety," starting out as a cheered-on underdog startup with a massive mission. Yet, it took only two years for their image to flip into a monstrous company aiming for the worst dictatorship in human history.

@kapilansh_twt It genuinely was amazing when it launched, but then they killed it by rate limiting the hell out of people's plans. I use it very rarely now (the constant permissions drive me mad). It's a shame. It had a lot of potential.

hot take Antigravity is a poor man's Codex Google made it free and still couldn't make it good best infrastructure on earth unlimited funding decades of AI research and they built the coding tool equivalent of a participation trophy