bindureddy's profile picture. CEO of @abacusai, the world’s first AI super assistant and general-purpose agent, Abacus AI Deep Agent, for enterprises and professionals. ex-GM, AWS and Google

Bindu Reddy

@bindureddy

CEO of @abacusai, the world’s first AI super assistant and general-purpose agent, Abacus AI Deep Agent, for enterprises and professionals. ex-GM, AWS and Google

AI and mass market robotics will make a huge difference in places like India which are in CODE RED Cities are breaking down from lack of infrastructure and massive overpopulation Traffic and public transport is in constant chaos, basic services like power and water are often…


It’s ridiculous for parents to pevent their kids from using AI Turn it around - give them all the AI models from day one and challenge them with 10x harder problems Give your kid superhuman powers 🚀


If AI were a human, it would be a 9 month old toddler Evolution took billions of years to make humans Relatively speaking we are barely getting started with AI In a decade, AI would have far surpassed all biological intelligence in every dimension


Google. Amazon, Cerebras, Groq and a dozen other companies are making chips In a decade, NVIDIA’s margins will go to zero But for the next few years, it will continue to be one of the most valuable companies in the world Nvidia is the Cisco of the dot com era


Goal For 2026 - Get a general purpose AI agent to build a deep vertical Fundamentally smart general agents should be able to do very specific things like be a kick ass legal assistant! In 2026, we will have a very deep legal assistant that is powered by general purpose agent…


The 2025 AI Crown Goes To 👑 OPUS 4.5 👑 Its head and shoulders above other models in real world agentic scenarios Runners up - Nano Banana Pro, Codex Max


LLMs have come a long way but they still have a long way to go in 2026 - be more aware of space-time - become more self aware - know when you are making things up - don’t apologize profusely The list goes on - but we are definitely on the path to AGI


2025 Was The Year Of The LLM 🚀 January - DeepSeek R1 Feb - Grok 3 Mar - Gemini 2.5 Pro April - Qwen 3 May- Sonnet 4.0 and Opus 4.0 Jul - Grok 4 Aug - GPT 5 Then Nov-Dec went absolutely nuclear- Grok 4.1, Gemini 3, Opus 4.5, and GPT-5.2. The acceleration will continue in…


LLMs are slowly automating white-collar work, even as haters continue to hate 🤷‍♀️ In 2026, they will - - create new features, test them, and release to prod - run vibe-marketing campaigns and run social media accounts - analyze data, generate reports, and come up with insights…


Meta is dropping two AI models - Avocado and Mango sometime next year It will literally be a HUGE MIRACLE if they come anywhere close to top models. They are almost 18 months behind which may as well be a decade in AI land


Definitely not slowing down because of the holiday break In fact, we are launching something pretty big next week. 🚀


Pro Tip - You are not really using AI if you only use small and cheap models. Most of the world never uses SOTA models like Codex Max or Opus 4.5 and is totally CLUELESS about what AI can really do.


The next big thing will be Grok 4.2, followed quickly after by GPT 5.5 OpenAI and Grok won't be slowing down because of the holiday break


LLMs have gotten extremely good at benchmark maxing The problem is that their performance on benchmarks often doesn’t translate to real world performance Far too many LLMs still fail at complex real world multi turn conversations while they look good on paper


Gemini 3.0 Flash Is An Insanely Big Deal - It Beats GPT 5.2 On LiveBench!! Plus, it is 10x cheaper 🤯🤯 This may be the BIGGEST launch of the year

bindureddy's tweet image. Gemini 3.0 Flash Is An Insanely Big Deal - It Beats GPT 5.2 On LiveBench!!

Plus, it is 10x cheaper 🤯🤯

This may be the BIGGEST launch of the year

Gemini Flash 3.0 - Intelligence Too Cheap To Meter? Gemini Flash series has been one of the best small models EVER. We have 100x usage on Flash compared to the Pro version. Flash 3.0 seems to be Google's best model yet. LiveBench results coming soon.

bindureddy's tweet image. Gemini Flash 3.0 - Intelligence Too Cheap To Meter?

Gemini Flash series has been one of the best small models EVER. 

We have 100x usage on Flash compared to the Pro version. Flash 3.0 seems to be Google's best model yet.

LiveBench results coming soon.

Gemini Flash 3.0 arrives tomorrow… TBH it’s the most used google given the speed and cost 🚀


OpenAI has a new image model - GPT Image 1.5 It's pretty good at instruction following and precision editing. First impression - may be comparable to Nano Banana Pro Both models will be available on ChatLLM today, so that you can compare and contrast for yourself


India is in “code red” Smog in large cities like Delhi is causing millions to get sick. Schools are closed! The planet is in a precarious place as we continue to ignore the impact of climate change 😢


Loading...

Something went wrong.


Something went wrong.