AI agents · How Founders Use AI
Should I build one agent or a team of agents working together?
Start with one. The famous industry debate (Cognition's "Don't Build Multi-Agents" vs Anthropic's multi-agent research system) actually agrees on the fundamentals: multiple agents working from partial context make conflicting decisions, and multi-agent setups burn roughly 15x the tokens. Cognition's 2026 follow-up narrows the rule: extra agents can research or review, as long as only one writes. Go multi-agent only for parallel, read-heavy work like broad research, with one lead agent delegating to sub-agents.
On the AI map, 29 October 2026 Decision modelsCoding agentsAgents that do tasks for youAI browsersNo-code automation with AI steps
Go deeper18 sources
-
Don't Build Multi-Agents
Walden Yan (Cognition) · Article · Jun 2025
The contrarian essay from the Devin team that reset the industry's defaults.
-
Walden Yan: principles for building agents
Walden Yan · Thread · Jun 2025
The thread version of the argument, with the mistakes he keeps seeing.
-
Walden Yan: a year later, what changed
Walden Yan · Thread · 2026
The honest update: some multi-agent setups now work, most sexy ideas still don't.
-
Multi-Agents: What's Actually Working
Cognition · Article · 2026
The follow-up naming the specific multi-agent patterns that survived production.
-
How we built our multi-agent research system
Anthropic · Article · Jun 2025
The other side of the debate: a 90% performance gain, at 15x the tokens.
-
When to use multi-agent systems (and when not to)
Anthropic · Article · 2025
Anthropic's own decision rules for graduating beyond a single agent.
-
Single vs Multi-Agent System?
Philipp Schmid (Google DeepMind) · Article · Jun 2025
Shows why Cognition and Anthropic are both right about different problems.
All 18 sources
-
How Anthropic Built a Multi-Agent Research System
ByteByteGo · Article · 2025
The architecture broken down with diagrams a non-researcher can follow.
-
Why Do Multi-Agent LLM Systems Fail?
UC Berkeley (MAST) · Article · Mar 2025
200+ analyzed tasks and 14 failure modes, the evidence behind the caution.
-
MAST: Multi-Agent System Failure Taxonomy
UC Berkeley Sky Computing Lab · Tool · 2025
The open dataset and annotator to diagnose your own multi-agent failures.
-
MAST project page
UC Berkeley Sky Computing Lab · Article · 2025
The digestible summary of where multi-agent systems break.
-
AI Leaders Clash Over Agent Architecture
CTOL Digital · Article · 2025
A neutral recap of the whole debate if you want it in ten minutes.
-
Multi-Agent vs Single Agent: When It's Worth the Cost
Cognilium · Article · 2026
Frames the choice as a coordination-cost calculation, which is the right lens.
-
The Art of Multi-Agent Collaboration: Reflections on Anthropic's Research System
LLM Multi Agents · Article · 2025
Extracts the engineering lessons hiding inside Anthropic's post.
-
Why do multi-agent LLM systems fail (and how to fix)
Future AGI · Article · 2026
Maps each MAST failure mode to a practical fix.
-
Why Do Multi-Agent LLM Systems Fail? (explainer)
Anna Alexandra Grigoryan · Article · 2025
The paper's findings retold for people who won't read a paper.
-
Anthropic's multi-agent research system (LLMOps case study)
ZenML · Article · 2025
The case study written for someone deciding whether to copy the pattern.
-
Harrison Chase on Deep Agents, LangSmith, and Earning Trust
NVIDIA AI Podcast (Harrison Chase) · Video · 2026
The emerging alternative: one deep long-running agent instead of many shallow ones.
Watch it here
