MM

[MM @0xmarma]

@bindureddy I wonder what use case drove this demand?

Gemini 2.5 Pro is great - but tool use is sub par even with the larger 1M, sooner to be 2M context windows.

o3 mini not exactly great in that area either.

Anthropic still the kind in this area - especially coding.

[MM @0xmarma]

OpenAI says that it plans to release both o3 and a next-generation successor, o4-mini, in weeks, following by GPT-5 a few months after.
.
The read is GPT-5 delay, and o3 to fill the release gap.
https://t.co/dtbWKl9WFQ

[Min Choi @minchoi]

ChatGPT GPT-4o Image Generation is breaking the Internet.

People can't believe it's 100% AI, it's insanely good.

10 wild examples:

1. Beachside selfie (prompt in ALT)

[MM @0xmarma]

Another great news from @AnthropicAI ! This @databricks partnership has a lot of upside for both:

1. More data for training/post-training for Anthropic - they just release search earlier as well.

2. More Enterprise Data and AI capabilities (60%+ of Fortune 500, 10k org worldwide)

3. Databricks data is pretty much plug-n-play - very low barrier of entry

4. Safety and Governance already in place

5. Competitive Edge - Snowflake has Claude 3.5 in Cortex AI integrated already, they do their own hosting though.

#databricks #anthropic

source: https://t.co/QoouukrJPF

[MM @0xmarma]

@bindureddy if you haven't tried it - so much better - i'm so looking forward to the APIs

[MM @0xmarma]

5/5 - Impressed w/ Gemini 2.5 Pro!

Feels like a shift to reasoning AI for devs, not just incremental.

Available now (Experimental) in AI Studio / Advanced users.

Eager to test on real projects!

#ArtificialIntelligence #FutureOfCode #GoogleGemini

https://t.co/klVFicEykr

[MM @0xmarma]

1/5 - Google's Gemini 2.5 Pro is here!

Big step up for devs.

Called their "most intelligent AI," #1 on LMArena.

Focus seems on how it thinks, not just benchmarks. Promising stuff.

Highlights in 🧵👇

#AI #Gemini #DevTools

https://t.co/AhOW56WaxQ

[MM @0xmarma]

4/5 - Better reasoning is key! Scores 18.8% on Humanity's Last Exam.

Helps me with complex logic bugs, system design from reqs.

Native multimodality (text/code/img/video) is cool too - analyze visual bug reports? 👇

🤔 #GeminiPro #AIReasoning

[MM @0xmarma]

@bindureddy Information gathering is table stakes now - i wonder what makes this AI understand the audience better - which is the key in this agentic world now.

[MM @0xmarma]

Manus, Operator, Computer Use signal a shift: AI as co-worker, not just assistant.

Tool use unlocks complex tasks - think real-time decisions, seamless workflows.

This is agentic AI’s promise.

Are we ready for what’s next?

https://t.co/5ZVq3qYFua

[MM @0xmarma]

AI’s next frontier?

Computer/tool use. @OpenAI's Operator dropped on Jan 23, and Manus AI (Mar 6) to the punch.

This isn’t just about timing—it’s about agentic AI taking control, automating tasks, and reshaping how we work.

Let’s dive into why.

🧵

https://t.co/CdjFzhJ7ue

[MM @0xmarma]

From Perceptron (1950s) to Operator (2025), AI’s evolution is about agency.

Finance? 77% of institutions use AI for portfolios. Agentic AI isn’t coming - it’s here, wielding tools to cut errors and boost productivity.

The future is autonomous.

https://t.co/fl4iq578UR

[MM @0xmarma]

@blader @blader what product/industry sparked this comment?

[MM @0xmarma]

@robgrondel @bindureddy it’s not a base llm model - ai agent on steroids if anything

[MM @0xmarma]

@bindureddy I don’t get the labeling exercise here - why does it matter if it’s a good open source model

Plus what makes the difference is how it’s used - ie in your example of Sonnet 3.7 / 3.7 thinking being best for coder - which I 10000% agree

[MM @0xmarma]

@_catwu @grok why wouldn't Claude Code offer an IDE?