The BasePodcastThe Cognitive Revolution92 min
Agents That Think Together: Cisco's Vijoy Pandey on Semantic Alignment, Scope, Accord & Concord
In a 92-minute interview with Nathan Labenz, Cisco's Vijoy Pandey describes multi-agent systems running in production, reports Cisco's measurements of silent coordination failures among agents, and presents three open-source tools, Scope, Accord and Concord, meant to detect and reduce them.
Why it matters
The same thing, for different readers.
What it means for you
Cisco's Vijoy Pandey says agent systems can look healthy on dashboards while completing none of the task, so AI assistants acting for people may fail without any visible warning.
For researchers
Pandey reports Cisco measured task accuracy in five-agent systems falling from 99 to 16 percent after three coordination failures, a company result not yet independently checked.
The question it raises
Pandey argues humans already live under systems they cannot fully grasp, such as high-frequency trading, and that records, monitoring and human judgment can keep such systems accountable.
In their own words
Every quotation was found, word for word, in the text we saved from the source. That shows the words are there; it does not make them true, or prove the speaker meant them as quoted.
From our own transcript of the episode’s audio, made by machine. Speaker names come from what is said, not from a voice match.
“We saw that 100% of chat dev sessions had semantic failures and cognitive misalignment failures.”
“With three cognitive failures, which is not a lot, the accuracy of the multi-agent system dropped to 16%.”
“the system was 66% up, which was defined as working perfectly, but it had 0% task completion”
“right now I believe that that verifier or that person providing the judgment is actually a person.”
“if it means building these latent space protocols where agents can talk to each other not in natural language but with higher dimensionality. Let's build those things.”
How the finding was reached
In our words, from the source. Weigh the finding by how it was made.
- What was done
- Cisco analysed public traces of ChatDev sessions and its own deployments of simple five-agent systems, counting semantic and cognitive misalignment failures against task accuracy.
- What it found
- Pandey says every ChatDev session showed such failures; accuracy of a five-agent system fell from 99% to 98% with one failure, to 81% with two and to 16% with three; a travel-booking system was 66% up with 0% task completion.
- What it does not show
- Spoken figures from a company executive, with no methods or sample sizes given in the conversation; Cisco is also releasing the tools it describes.
- Status
- The company’s own report about its own work
The record
- Questions
- Can AI systems act beyond what people intended or allowed? · Are AI companies putting stronger safeguards into practice? · How fast is AI advancing, and where is it heading?
- People
- Vijoy Pandey (Speaks, SVP and GM, Outshift by Cisco) · Nathan Labenz (Speaks, Host, The Cognitive Revolution)
- Also covered by
- The Cognitive Revolution (YouTube)
- What we read
- We read the whole source.
- Found by
- The Radar, run of November 12, 2026, via
pod-the-cognitive-revolution(tier 1: the source’s own channel or an official record) - Cite this
- The Cognitive Revolution, “Agents That Think Together: Cisco's Vijoy Pandey on Semantic Alignment, Scope, Accord & Concord”, November 11, 2026. Entry in the Ctrl AI Base, https://ctrlai.com/base/cognitive-revolution-2026-11-11-cisco-pandey-agents-think-together