Yogita C

Would like to see more people with positive attitude.

[Yogita C @yogitac08]

@IndependentEco Microsoft 365 Copilot leverages a multi-model architecture at the enterprise level, incorporating models from both OpenAI and Anthropic.

[Yogita C @yogitac08]

Sunday thought: the best AI skill in 2026 isn’t prompting. It’s knowing which tasks to never hand to an agent. Do you agree?

[Yogita C @yogitac08]

Which do you trust more with real work today?

[Yogita C @yogitac08]

@StephenJohnston @lucas_builds I second that. Fable 5.1 is good with detecting and fixing too, but might use more tokens so handover to Opus 5.5.

[Yogita C @yogitac08]

In 2026 the model matters less than the permissions you give it. Everyone compares benchmarks, and nobody compares what the agent is allowed to touch. What do you think?

[Yogita C @yogitac08]

Google’s Gemini 4 Argon has landed, taking the top spot on 13 of 18 industry benchmarks.
The catch? It’s not fully public yet. Here is what you need to know:
Initial rollout is limited to select cybersecurity partners under the Fairwind Program.
The Cost: Early API access is priced at $2/M input and $10/M output tokens (expected to double later).
Would you switch to Gemini?

[Yogita C @yogitac08]

Small models are getting cheap enough to run on every request. Haiku 5.5 is $0.10 per 1M input tokens, and the price war is now in the small-model tier. Do you still pick the biggest model by default?

[Yogita C @yogitac08]

@Olking07 The pivot from "AI is stealing all our jobs" to "AI is creating new opportunities" happened fast. Let’s just hope the optimists are right on this one.

[Yogita C @yogitac08]

How can we trust OpenAI when they fire the very safety researchers who voice concerns about their technology?

[Yogita C @yogitac08]

OpenAI posted hundreds of AI-generated math proofs at once. Mathematicians are split between ‘breathtaking’ and furious. Peer review was built for a few papers at a time. Who checks hundreds? What is the accuracy percentage of these solutions?

[Yogita C @yogitac08]

@pelloiafilippo Claude and Gemini

[Yogita C @yogitac08]

Thought of following the trend and create a "A Day in life of Claude". I am really pleased on the progress Sonnet 5.5 has made over previous models. The music and story telling is a major development.
Drop in what you have made with the latest Claude models. Lets see your creative side. :)

[Yogita C @yogitac08]

@jimmy_dj_c Congratulations on your launch. Failing is a journey to success. Wishing you great success.

[Yogita C @yogitac08]

Muse hit 5M downloads in under a month, and it's also under fire for privacy. Would you give a personal agent access to your messages if it saved you hours a week, knowing that the AI companies can use your data?

[Yogita C @yogitac08]

@EbiThinks Isn't Meta developing an "AI CEO agent" and an AI clone of Zuckerberg designed to act as an executive co-pilot. So yeah it might be possible in the near future but might not be a completely autonomous agent.

[Yogita C @yogitac08]

@alwayspriyesh Well they launched their own marketplace which is new. The personal assistant Dot looks promising too.

[Yogita C @yogitac08]

My read of OpenAI DevDay: it wasn't about a smarter model (more like a cheaper and faster one). It was about agents, distribution and price.

Everything announced, on one card

The open question: as agents get more autonomy, who checks their work?

[Yogita C @yogitac08]

OpenAI DevDay is tomorrow. Fortune reports OpenAI paused major launches for two weeks to ship a dozen plus products in one go. What's your one prediction? Mine: an agent platform, not another model.
Reply and I'll check back after the keynote tomorrow.

[Yogita C @yogitac08]

@full_kelly_ @claudeai Exactly my thoughts, is this slowing down?

[Yogita C @yogitac08]

@elonmusk Is it getting worse, or have these agents become so powerful that even OpenAI can no longer control them? Who knows the truth..