The A.I. Daily
Bindu Reddy@bindureddy
Confirmed! Argon absolutely nails data analysis and excel tasks Just flipped it on in production An insanely huge quality jump compared to Opus, and it's cheaper too ♥️
Bindu Reddy
170 storiesBindu Reddy is an AI commentator who posts on X about model capabilities, pricing and where the industry is heading. Recent posts focus on Google's newly announced Carbon model, including his LiveBench expectations, pricing speculation and a rumored OpenAI model called Bel, plus his purchase of Google shares. He also covers Argon's weak instruction-following, Chinese models distilling Astra, Anthropic's reliance on API revenue, and AI moving from coding to physics.
Bindu Reddy@bindureddy
AI market in the next 6 months - open-source models will get better - token prices continue to drop - frontier Labs will be forced to cut prices more - fable and astra forced to cut prices to $30/1M token - 80% of all tasks are done by cheap models
Bindu Reddy@bindureddy
🚨 DeepSeek Flash Is The Most Cost-Efficient Model In The World - decent performance - only $0.13 / task - excellent reasoning and coding Open router stats show the same thing. This is the most used model in the world in terms of tokens
Bindu Reddy@bindureddy
Google is definitely back with Argon A couple more iterations and they will almost certainly catch up to OpenAI and Anthropic They have a unique advantage right now because the frontier is stalled for “safety reasons”
Bindu Reddy@bindureddy
Gemini 4.0 Argon is #1 in the WORLD at - data analysis - reasoning Yes, better than Fable 5.1 and Astra 6 in both these categories
Bindu Reddy@bindureddy
Fable 5.5 or Astra 6.1? - Fable is probably going to drop before Asta - My guess is both of them will cut prices - Lots of room for frontier models to improve on hard code and data analysis Likely drop in the next two weeks. Google should drop Carbon as well
Bindu Reddy@bindureddy
🚨 Gemini Argon 4 Falls Behind Fable And Astra, But Is Ahead of Open Source - Argon 4.0 falls short of GPT 6.1 Sol - ahead of open-source (Deepseek Flash) - much better than Muse Wonder - Muse is marginally better than DeepSeek This is all on our new benc…
Bindu Reddy@bindureddy
🚨 Self-Improving Agents Maximize Your Revenue - set a goal - agent iterates on a regular basis - each run improves metrics - hit your goal over time Get more conversions, make viral posts and close more deals with swamrs of self-improving agents
Bindu Reddy@bindureddy
Can confirm! Muse Wonder is nowhere near Fable or Astra level This is all getting pretty tiresome. We will publish our new bench shortly with all the details
Bindu Reddy@bindureddy
Super Excited For Our New Agentic Long Horizon Benchmark - LiveBench 2.0 We have hundreds of benchmarks, but most of them are just plain broken and bench-maxxed. We have worked over the last several weeks to fix this - launching whenever Argon launches - ref…
Bindu Reddy@bindureddy
SuperIntelligence is going to solve mass-market robotics shortly Get swarms of agents to run billions of simulations of a robot model in a physical world Pick the best robot models and try them in the real world to make a mass-market robot A robot that aces p…
Bindu Reddy@bindureddy
Fable 5.1 rocks at coding but sucks BIG TIME on data analysis In sharp contrast, Astra does remarkably well on complex data analysis Argon will be better than both in data analysis
Bindu Reddy@bindureddy
October Upset - Gemini Argon is sooo fantastic that it beats both Fable and Astra It's all 5x cheaper than both models It will automatically force a price cut on Astra and Fable Probabilty of this happening - 55%
Bindu Reddy@bindureddy
Google’s employees think that their internal model Carbon is very good Better than Gemini Pro 4 Argon that hasn’t launched yet! Argon became a legacy model even before it dropped 😲 Google can still pull ahead if they hustle and release Carbon in a week Just…
Bindu Reddy@bindureddy
Frontier models hard at work and moving at light speed!! It took OpenAI and Microsoft a couple of weeks to replicate Jev Humans no longer create or replicate technology, AI does. It simply does it 10x faster and better 🚀
Bindu Reddy@bindureddy
Experimenting with Astra vs Opus 5.5 with some of our biggest customers Generally they prefer Astra on hard core coding if the model is anonymous and the price is not revealed! Astra wins 80% of the time and results in very strong engagement
Bindu Reddy@bindureddy
OpenAI seems to be having a “safety crisis” Their model dev appears paused while Anthropic and Google are forging ahead Unclear what is happening within the company but I bet it is some version of office politics under the guise of “safety”
Bindu Reddy@bindureddy
Lower prices are simply coming with worse performance 😲 Frontier labs are cutting prices aggressively but they are also nerfing the models to make them more efficient Luna 5.6 is still the best variant because Luna 6.1 performs worse
Bindu Reddy@bindureddy
Fable and Astra still have a lot of room for improvement - they make random assumptions - don’t think holistically - extremely bad at complex UX - no sense of time or AI effort We are at least two generations away from the Super Intelligence! Plenty of time f…
Bindu Reddy@bindureddy
Keep Faith In DeepSeek and Open-Source AI The Chinese companies are the only ones that are banned from wrapping Anthropic or OpenAI models into random features They actually have to deliver extremely performant and efficient models to succeed So far, they hav…
Bindu Reddy@bindureddy
A math professor from NYU is accusing LLMs of “stealing” everyone’s research ideas Simply because OpenAI trained on public data! Pretty astounding statement Where does he get his ideas from? Thin air or by reading prior art! 🤯
Bindu Reddy@bindureddy
Google seems to be blowing it all up again 😭 The excitement about Gemini 4 Argon is fading… This is tragic especially if it’s a decent model They should make it generally available ASAP!
Bindu Reddy@bindureddy
Applications of Super Intelligence - solve all math problems - crack mass market robotics - build complex systems autonomously - cure aging and literally all diseases - invent near-infinite energy - invent smaller LLMs like Sol and Haiku for humans to play wi…
Bindu Reddy@bindureddy
Haiku 5.5 Is A Cheap Model That Underperforms In Its Class - MUCH worse than DeepSeek - worse than Luna as well Anthropic is better off launching Fable 5.5 🤷♀️
Bindu Reddy@bindureddy
TBH, we don't need one more super cheap model DeepSeek, Luna, Haiku, GLM, blah blah blah How about Fable 5.5 or Astra 6.5?
Bindu Reddy@bindureddy
Hearing rumors that Haiku 5.5 may slash prices by almost 10x Yes, 10x!! Will compete with DeepSeek and Luna Common man's intelligence has officially become too cheap to meter
Bindu Reddy@bindureddy
Stop focusing on boring personal agents that are apparently gonna book tickets 🙄 Who cares? Those booking sites block these AI agents anyway Meanwhile, SuperIntelligence continues to self-improve and is locked inside the walls of a duopoly The rest of the wo…
Bindu Reddy@bindureddy
The battle over the coming weeks will be for third place in the race towards super intelligence. Gemini 4.0 drops and is in the Fable or Astra league Or GLM 5.5 is awesome and puts open source ahead of Gemini
Bindu Reddy@bindureddy
Path to the age of abundance - super intelligence solves robotics - robots rapidly experiment in real world labs - super intelligence + robots invent near-infinite energy Problem solved! Resource constraints no longer exist Beginning of Utopia 🔥
Bindu Reddy@bindureddy
OpenAI just dropped solutions to a whole bunch of hard Math problems The AI solved more problems than what humans could solve in a decade Expect AI to ace all science problems very soon!