Small Teaser of our IO (Industrial Operator) 2.0 An AI system for Industrial Control Rooms. Version 2 going to production in early September. Version 3.0 (2025 Q4) in the making will use a new "Omniscient feature" we are developing based on hierarchical Realtime agents, where

it's incredible how much smarter gpt4.5 is than gpt-5. Not as capable, but surely smarter. if gemini3pro is as smart as gpt4.5 and as capable as gpt5... i need it!

now run FBAM-style long episodes and at the same time deploy ReasoningBank to harvest strategies and the param pressure becomes much smaller??

Is anyone using Chatgpt Pulse? and if yes, how useful so far?

The next thing to Agents is "Autonomous Reasoning Systems" (ARS?) and we, and others, are already building this. An ARS is not a chatbot / workflow activated by a human prompt or even a programmatic trigger. It is a system that understands its role and creates its own

There is only 1 thing i want from open ai to build agrntic systems: better LLMs

the difficulty and the sauce of an agentic application is not on knowing how to call the api,connect multiple conversations/agents, use tools or integrate MCP. those are all fairly basic technical tasks. the sauce is in the distribution of roles and intsuctions, in the specific

its October 2025 and openai prod team still don't know how to make an event-based agent demo (or product!) other than a user just prompting. Still in chatbot mindset. u better keep thinking...

let me grab some air to laugh really out loud... so after 3 years you develop a framework create an agent , no that it prevents you to go down on sales 30% but that you have to go and ask yourself when u notice u r 30% down hahahahahahahahahahhahaha

now I understand Ilya...

It's going to be SO BORING to see the 100s of posts showing "what i did in some mins with the agent builder"... I'll take a break and keep thinking.

Basically, if you really need that, I don't know what you've been doing the past 3 years...

vendor locking your AI workflow at this stage doesn't sound like a smart approach. This reminds me much those no-code frameworks made just to lower the bar to develop SIMPLE LLM-based apps. that's my take. not thrilled.

The overwhelming response to the @claudeai campaign is more than merch and coffee. It feels like a brief collective exhale. A reminder that we still long for the true, the good, and the beautiful.

There is no new AI leap needed, just expected inference/compute optimizations. Any AI performance leap on the way would only accelerate the transition.

Between the end of 2026 and mid 2027, Reasoning Systems will not only control large industrial plants, utilities and production lines fully autonomously, but they will also be able to do many other segment jobs.

A system like this scaled to e.g. a full power plant might consume 100s of aggregated ktok/sec; a cost anywhere in the 5-10k USD / day at today's pricing... but we are not building for today

A Reasoning System does not wait for a user to prompt it but its driven by events and processes embedded in the hierarchical network of interconnected agents

A Reasoning System should be autonomous and continuously debug and optimize your infrastructure, facility, plant or production line.

Our Industrial Operator (IO) will become the first "Industrial Reasoning System". A new category we created for it just to distinguish from the myriad of copilot, agents and chatbots around.

many talking about Anthropic's campaign, the colours, the vibes... I think it is more about the bloody message, isn't it? about thinking vs more brain-rotting? I don't care about the color, vibes or style, make an ugly af hoodie saying "Keep thinking. Building cancer-curing