Yes, you can. Unfortunately, OpenAI doesn’t make their real reasoning traces available, so it’s really hard to know if you’re doing it right with their models. Other models are different, you can see the actual reasoning so you can tell if it’s being steered properly!
@mattshumer_
-
Composio Launches Major Tool Use Advancement for AI Agents
By
–
This is a huge step forward for tool use. Congrats to the @composio team on a great launch! Check it out here:
-
Unified MCP Rube Routes Queries Across AI Tools
By
–
The Unified MCP is finally here.
— Matt Shumer (@mattshumer_) 21 août 2025
One MCP to rule them all!
Rube is the first MCP I’ve seen that routes your query to the right MCPs, in real-time.
Integrated into Cursor, Cline, Claude, etc.
Watch it analyze thousands of tweets on ChatGPT + organize by pain point in a sheet. pic.twitter.com/3WLrJs4fEyThe Unified MCP is finally here. One MCP to rule them all! Rube is the first MCP I’ve seen that routes your query to the right MCPs, in real-time. Integrated into Cursor, Cline, Claude, etc. Watch it analyze thousands of tweets on ChatGPT + organize by pain point in a sheet.
-
GPT-5 Thinking paradigm represents frontier AI progress
By
–
It's totally fair! The thinking paradigm is part of the progress we've seen! GPT-5 w/o thinking is not frontier, GPT-5 Thinking/Pro is. That's like saying GPT-3 vs. GPT-4 isn't a fair comparison because a post-trained GPT-3 might be closer than vanilla GPT-3.
-
GPT-5 Coding Capabilities Far Exceed Original GPT-4 Performance
By
–
I get what you're saying, but do you remember what it was like to code with the original GPT-4? At best, you were getting a function or two out at a time, and more often than not, they were riddled with errors. Essentially zero agency. GPT-5 on the other hand…
-
GPT-3.5 and InstructGPT Performance Compared to GPT-4
By
–
GPT-3.5, and the only other one I can think of, InstructGPT, were a lot dumber than GPT-4. They were much closer to GPT-3 with some tuning + sanding on top. Back when InstructGPT came out, we actually kept most of our workflows on the OG GPT-3 as it worked better for us.
-
Major coding capability improvement discussed
By
–
I generally agree. Though I do feel like it's a pretty massive step up for coding, at least on my tasks.
-
Why GPT-5 Release Doesn’t Feel Revolutionary Progress
By
–
Everyone is missing the obvious reason GPT-5 doesn't feel like a huge leap: AI progress doesn't feel exponential anymore, but it's still moving just as fast as ever. From GPT-3 to GPT-4, we saw almost no significant model releases in between, so GPT-4 felt revolutionary when it
-
Recent AI Models Surpass GPT-4 Capabilities in Reasoning and Multimodality
By
–
Exactly. Recent releases have given us reasoning, adaptive thinking time, great multimodality, long context, etc. etc. So many things that we see today and are like "oh of course the new model has that", seems like everyone forgot that GPT-4 had none of it.
-
OpenAI dominance ends: dozens of incremental models flood market
By
–
No. In short, OpenAI was the only serious game in town from GPT-3 -> GPT-4 release. Post-GPT-4, there have been dozens (hundreds?) of incremental model releases from OpenAI, Anthropic, Google, etc. etc. GPT-3 -> GPT-4 felt so drastic because we didn't see much in-between