@noisyb0y1: OXFORD AND ANTHROPIC SPENT $6.7M AND 4 YEARS - AND FOUND WHY 90% OF AGENTS FAIL ON COMPLEX TASKS most agents store fact…
Summary
A joint Oxford-Anthropic study, costing $6.7M over 4 years, found that 90% of AI agents fail on complex tasks because they store facts but lose connections; using a graph-based approach improved task success by 42%, reduced unnecessary calls by 33%, and increased research accuracy by 39%.
View Cached Full Text
Cached at: 07/21/26, 06:40 AM
OXFORD AND ANTHROPIC SPENT $6.7M AND 4 YEARS - AND FOUND WHY 90% OF AGENTS FAIL ON COMPLEX TASKS
most agents store facts but lose connections - and those connections determine quality on complex tasks
8,750+ tasks: task success up 42% - unnecessary calls down 33% - research accuracy up 39%
the graph knows not just what it knows but how confident it is - no list can do that
an agent with fixed context degrades - an agent with a graph accumulates experience and gets better with every iteration
bookmark & read the article below - build your first Graph Engineering tomorrow
Similar Articles
@Sprytixl: STANFORD AND ANTHROPIC SPENT $3.1M TO PROVE YOUR AGENT PERFORMS 42% WORSE THAN IT SHOULD - AND FOUND THE FIX most agent…
Stanford and Anthropic research shows that AI agents with graph-based memory perform 42% better than those without, improving code accuracy by 36% and research by 45%.
@rohanpaul_ai: New Anthropic research shows AI agents may look brilliant at code, but in biology they can fail before the science star…
Anthropic research reveals that AI agents struggle with biology databases, producing highly variable answers for the same query (e.g., Ebola sequence counts ranging from 5 to 106 vs. expected 266), but adding a repeatable retrieval tool significantly improves consistency and accuracy.
@zodchiii: Three Anthropic engineers just spent 16 minutes on what makes AI agents actually succeed in production. If the people w…
Anthropic engineers share insights on making AI agents succeed in production, highlighting proven patterns from their work on Claude.
everyone's focused on whether their agent works. almost nobody asks if it's actually getting better over time
The article points out a common oversight in AI agent development: while most teams monitor task completion, few systems capture and feed failure patterns back into future runs to enable learning and improvement over time.
@0xNoryxx: STANFORD SPENT 2 YEARS AND 9,842 TASKS TO PROVE THAT LOOP ENGINEERING PLUS GRAPH MEMORY IS THE ONLY WAY TO BUILD AGENTS…
A Stanford study across 9,842 tasks demonstrates that combining loop engineering with graph memory produces AI agents that outperform standard ones by 38.6%, with 24.1% fewer unnecessary tool calls and 21.7% lower latency.