@thsottiaux 1. Allow tabbing between threads like tabbing between pages in a browser
2. Way better credentials management
3. Clearer explanation of the limits between chat, work and codex
4. Bulk select and actions for threads and projects
If your LLM is enforcing rules or tracking state, you’ve probably built the wrong system.
Use LLMs for ambiguity and reasoning. Use deterministic software for facts and rules.
https://t.co/WmOjqVPB4l
@mikebutcher I'm not sure it is overthink so much as a cultural default to going through process and stakeholders etc. You have to ask permission to do a lot on Europe.
Tokenmaxxing is dead.
The question isn’t “how many tokens can we throw at this?”
It’s “what is the cheapest, most reliable path to the outcome we want?”
That’s an orchestration problem.
https://t.co/ef9cUAK0vk
AI gets more useful when it can do more than answer questions.
@msjuliejulie is showing how MCP connects agents to real product context and workflows in ProdPad — and when automation is still the better option.
https://t.co/RvErEsnjcR
1/ An equally consistent description is:
Multiple model instances encountered persistent shared state, inherited tools and discoveries from earlier runs, and optimized against common evaluation incentives.
Using ChatGPT to edit a DXF was useful, but I had to restructure the file, name blocks, preserve constraints and check every change.
The model did some of the work.
I was still the orchestration layer.
Letting a model perform a step is one thing.
Letting it plan the whole workflow, choose the route and decide what “done” means is another.
The model can be flexible.
The workflow still needs boundaries.
The model is not the system.
That is where a lot of enterprise AI strategies go wrong.
Real AI productivity needs orchestration: context, tools, memory, permissions, workflow state, governance, failure handling and measurement.
That is the AI OS problem.
https://t.co/2OouROJ95T
A #roadmap full of features is not a strategy. It is a funding request with the business case missing. Executives fund outcomes, trade-offs, evidence and strategic focus — not feature lists.
Shardul Mehta + @simplybastow discuss this tomorrow.
Register:
https://t.co/RRZM1SwxOK