Most AI talk falls into 2 camps:
๐ค Maximalists โ โAGI is here, the revolution is unstoppable.โ
๐ Minimalists โ โAI is dumb, donโt worry.โ
Both miss the point.
In Rebooting AI (2019), Marcus & Davis showed how AI failed simple reasoning โ like telling a cozy fireplace from a house fire.
I re-ran their tests in 2025: ChatGPT got almost all right.
So whoโs right โ the skeptics or the believers?
Neither.
The real question: how do we govern and shape AI, now that itโs fast becoming an evolutionary force on society itself?
https://t.co/rb36C0yExL via @LinkedIn
I built a quick prototype: Argumenta ๐ ๏ธ
It uses AI to break down articles and surface sharp
questions that make you think deeper.
No login
Just paste an article
Get structure + a โSecond Takeโ
Still early. Scraping may break. Logic might miss. But the core ideaโs here.
Try it: https://t.co/ZD9UXeOJC2
RTs & feedback = gold.
@lovable
10/10
This is just the start.
Iโm refining the orchestration, testing more debates, and applying it to new domains.
If you're exploring multi-agent AI systems or LLMs for strategic reasoning โ let's connect.
#ArtificialIntelligence#LLMs#GenAI#Reasoning#AIagents #PromptEngineering #Debate
๐งต I made two AIs debate the future of Italyโs education system.
Not a Q&A. Not a prompt test.
A full, multi-round argument between LLMs, judged by other AIs.
The results were surprising โ and useful. Here's what happened ๐
3. I locked two LLMs into opposing roles:
- Agent A: argue *for*
- Agent B: argue *against*
Each round included arguments, rebuttals, and counter-rebuttals.
Then 4 separate LLMs acted as judges.
No summaries. No fluff. Just structured debate. (3/10)
9. Hereโs the full write-up with methodology, prompts, examples, and takeaways ๐
๐ https://t.co/IsnaSEiVz5
If youโre working on agentic AI, orchestration, or multi-agent reasoning โ Iโd love your take. (9/10)
8. I documented the full setup:
- Prompt orchestration
- Round-by-round structure
- Evaluation criteria
- What worked, what didnโt
- How this could be used in policy, strategy, or product reasoning (8/10)
7. I stayed in the loop โ but stepping back let me observe, not just react.
It felt less like using a tool, and more like building a system to think *with* me.
This wasnโt about better answers.
It was about sharper thinking. (7/10)
6. The biggest surprise?
Letting AIs debate each other gave me more insight than any 1:1 chat ever did.
It forced nuance.
It surfaced contradictions.
It challenged assumptions I didnโt think to question. (6/10)
5. The debate unfolded across 4 core topics:
- Relevance of Curriculum and Teaching Methods
- Effectiveness of Teacher Training
- University or Labor market readiness for Students
- Administrative Efficiency, and Capacity for Reform
Each round revealed different strengths โ and failures. (5/10)
4. One of the strongest moments?
Agent B hit hard with this:
โItaly has one of the highest youth unemployment rates in the EU and a NEET rate of 16% โ well above the EU average of 11.2%.โ
That stat reframed the entire debate. (4/10)
/5 We are builiding AI Academies? Why would anybody want to go to an academy that has no pedegree and no history? This is the longest bunch of nonensense I heard in a long time. If this is the EU, then the EU must go.
/4 It seems somebody who has just read some book on AI and is regurgitating some random words like the worst hallucinating LLM. We are building cloud capacity for AI? Who wil own this? Which companies?