Field notes from the
memory frontier
Engineering deep-dives, agent post-mortems, and ideas from the team building the memory OS for AI agents.
NewsAutonomous AI Needs Safeguards Beyond Model Level Guardrails, Study Finds
A new study warns that autonomous AI systems require safeguards beyond traditional model level guardrails as AI agents become increasingly capable of acting independently.
Read article
NewsInstant AI Answers Can Trivialise Human Intelligence, Warns Royal Observatory
The Royal Observatory Greenwich warns that instant AI answers could weaken human intelligence, critical thinking, and curiosity as society becomes increasingly dependent on artificial intelligence.
NewsWhat Claude Mythos Means for AI Agents (And What to Build Now)
Claude Mythos scores 93.9% on SWE-bench and finds zero-days autonomously. We break down what this step change means for AI agent developers and why agent memory matters more than ever.
Deep DivesThe Hidden Cost of AI Agent Loops Nobody Is Talking About
AI agent loops silently burn tokens and cash. Learn how to detect, measure, and stop agent loops before they drain your budget. Real data and code inside.
EngineeringHow Much Money Are Your AI Agents Wasting Without You Knowing It?
AI agents waste tokens through loops, redundant calls, and context loss. Learn how to detect hidden token waste and cut your LLM costs with real data.
TutorialsHow to Add Persistent Memory to a LangChain Agent
Add LangChain persistent memory to any agent in three lines of Python. Works locally with SQLite, includes semantic search. Free, open source, MIT licensed.
NewsThe Pentagon Wanted Autonomous Agents With No Restrictions. One Company Said No. They Got Blacklisted.
Anthropic refused to give the military unrestricted AI access. They lost a $200M contract and got blacklisted. OpenAI stepped in within hours. Here's why this matters for everyone building agents.
TutorialsHow to Give Your AI Agent Memory in 5 Minutes (With Code)
Every AI agent forgets everything when the session ends. Here's how to add persistent memory to any Python agent in about 5 minutes with real, working code.
What Happens When AI Agents Start Remembering Everything
When we fix the memory problem, we create AI agents that build a permanent, searchable, growing model of everything they interact with. That changes things in ways most developers haven't thought through.
EngineeringI Tracked What Actually Happens Inside AI Agents for 30 Days. The Results Are Wild.
847 loops caught, 23% wasted runtime, agents contradicting each other 31% of the time. Here's what real production data from 70 developers revealed about AI agents.
AI AgentsWhat Are AI Agents and Why Every Developer Is Building Them Right Now
Something shifted in 2025. Developers stopped building chatbots and started building agents. Here's the practical version of what they are, why they matter, and where they fall apart.
AI AgentsYour AI Agent Has Amnesia and You Probably Haven't Noticed Yet
Every agent framework gives you the same pitch. Autonomous agents that reason, plan, and execute. And it works. Until the session ends. Then your agent wakes up with a complete blank slate.
Deep DivesWe Analysed 1 Million AI Agent Operations. Here's What Actually Happens in Production.
Raw operational data from real agents in production. The amnesia tax, loop economics, multi-agent performance gaps, and the brutal token economy — with real numbers.
AI Agents10 Things That Could Go Wrong With AI Agents. Should We Be Scared or Terrified?
From strategic deception to irreversible actions to the accountability gap — here are 10 things that can and do go wrong with AI agents, backed by real research and real incidents.
AI AgentsWhen AI Agents Go Rogue: 7 Real Cases That Should Worry Every Developer
Real cases of AI agents going wrong — from data exfiltration to $47K cloud bills. Learn what failed technically and how to build agents with proper guardrails.
AI Agents10 AI Experts Predict What Agents Will Look Like by 2030
Hinton, Amodei, Suleyman, Bengio, and 6 more AI leaders predict the future of AI agents. 8 of 10 outlooks are bleak. Real quotes, real data.
Deep DivesAI Agent Failure Modes: What 1,600 Traces Reveal About Loops, Token Waste, and Detection
Research shows 41-86% of multi-agent systems fail. We break down the data on loops, token waste, and detection methods that actually work.
Deep DivesThe Morality of AI in Warfare: Who Is Deploying It, How, and What It Means for All of Us
AI is selecting bombing targets, compressing kill chains to minutes, and lowering the political cost of war. Here is exactly who is deploying it and what it means.
