2026
- Which Agent Platform? Bedrock Managed Agents, AgentCore Harness, or Runtime
TL;DR: Amazon Bedrock Managed Agents, powered by OpenAI, went into preview on September 29, and it is easy to read the launch as a choice between two products: Managed Agents or AgentCore. Read …
- Weekly Review — Oct 26–Nov 1, 2026
TL;DR: Four posts this week, and three of them are about what an agent’s safeguards actually rest on. I ran my own agent setup through Shostack’s four threat-modeling questions. Most of my rules came …
- Wrong, Not Broken: An Agent Evaluator Is Only as Good as Its Reference
TL;DR: Matt Wood’s “Wrong, not broken” describes a refund agent that answers in 600 ms with zero errors and quotes $40 where the policy says $140, because it retrieved last quarter’s document [1]. …
- Weekly Review — Oct 19–25, 2026
TL;DR: Five posts this week, and four of them ask the same question about agents: what does the green check actually rest on? An agent paying per inference call is held back by a spending cap that …
- Your Agent's Notes Are the Closest Thing It Has to Training Data
TL;DR: A new paper from Meta Superintelligence Labs, UC Berkeley and others measures how well frontier-model agents improve when their weights stay frozen and everything they learn goes into files: …
- Your Agent Needs to Remember What Passed
TL;DR: Ten pieces, 1% odds each. Guess all ten at once and you need about 10^20 tries. Keep each piece a check confirms and you need about 1,000. In agent loops, the model raises the odds that a …
- Who Referees the AI?
TL;DR: In July 2026 a purported disproof of the Collatz conjecture got through Lean’s official proof checker, and through an independent second checker as well. It was wrong. The submission exploited …
- Weekly Review — Oct 12–18, 2026
TL;DR: Five posts this week, and in most of them the part that decides the outcome sits outside the frame you were looking at. The sender stares at the silence, while the real cause is the receiver’s …
- The Matryoshka of Humans in the Loop
TL;DR: Amazon Mechanical Turk closed on 30 September 2026. Its own tagline, “artificial artificial intelligence,” is a useful lens for what came after it, because the human-in-the-loop never left the …
- Weekly Review — Oct 5–11, 2026
TL;DR: Six posts this week, and most of them ask the same question from different seats: who carries the consequence of a decision, and can they see it? A line in a reorg plan that the planners can …
- GLM 5.3 on Amazon Bedrock: An Open-Weight Coding Model Behind a Managed API, Tested
TL;DR: GLM 5.3 from Z.ai is a mixture-of-experts coding model with roughly 750B total and about 40B active parameters, a 1M-token context and always-on reasoning, and it is now on Amazon Bedrock for …
- More Than Technology: How Agentic AI Changes Minds, Culture, and the Way We Build
TL;DR: Once usable agentic technology is available, the bottleneck often lies in how people, culture and ways of working change. For individuals, the core skill shifts to defining good, giving …
- Can You Trust a Swarm's Reasoning?
TL;DR: When a collective of agents coordinates internally and a deployment surfaces only its final broadcast to you (a design I am positing to make the trust question concrete, not something the …
- Four Talks, Four Rooms, One Question: Who Clicks Last?
dri TL;DR: In one week in Munich I gave four talks to four very different audiences: women building their careers at SAP and Amazon, observability engineers at Grafana, advertising agencies at a Wiesn …
- When Your Benchmark Runs for 88 Hours
TL;DR: On the Dwarkesh podcast, Noam Brown of OpenAI described a system of roughly 10,000 agents that spent 130 billion tokens over 88 hours on a single problem. This is his account, not independently …
- What If We Didn't Teach Agents to Work Like Us?
TL;DR: In Brown’s account on a recent podcast (his description of OpenAI’s system, not independently confirmed), a collective of agents given a bare message-another-agent primitive works out …
- When the Model Becomes an Org
TL;DR: In September 2026, Noam Brown of OpenAI said on a podcast (not independently confirmed) that a system of roughly 10,000 agents spent 130 billion tokens over 88 hours on one of the Millennium …
- Foundations of Agent Authorization
TL;DR: Agent authorization looks like a scary new frontier, but almost all of it is OAuth 2.0 and OpenID Connect, the same flows that already sign you into web apps. Once you can tell apart the three …
- Which of Your Work Is Actually Verifiable?
🎬 Also available as a blog walkthrough video. TL;DR: Two well-argued pieces already say the same thing about the AI era: AI takes over the work that is cheap to verify, and human value moves to the …
- Every Catalog Wants to Be an MCP Server
TL;DR: AWS just shipped an open-source MCP server for the Registry of Open Data — 1,100+ datasets from 400+ organizations (NASA, NOAA, NIH, the Allen Institute), now discoverable by asking your AI …
- Your AI Pair Programmer Is Playing Single-Player
TL;DR: I run what looks like a small team of agents — a backlog of delegated tasks, subagents for research and drafting, a whole operating model for handing work to machines. It feels collaborative. …
- When the SOC Analyst Is an Agent
TL;DR: Amazon GuardDuty now ships an AI investigation agent (in preview) that takes a raw finding, pulls 90 days of related activity, correlates it against threat intel and a knowledge graph, and …
- The AI That Only Watches
TL;DR: The whole industry is racing to give AI more permissions: more tools, more autonomy, more write access. I went the other way this month and noticed something uncomfortable: the assistant that …
- Weekly Review — July 27-August 2, 2026
TL;DR: Three posts about where the human stays in charge as agents do more of the work. “Laws & Disorder” closes with Tesler’s Law: you can’t delete a system’s complexity, only choose who owns each …
- Understanding Is the New Bottleneck — and Your AI Agent Can Fix It
TL;DR: In AI coding the hard part is no longer writing the code, it is understanding the code the agent wrote well enough to keep steering. Geoffrey Litt’s framing is that the human job shifts from …
- Generate-and-Verify: What P vs NP Teaches Us About AI Reasoning
TL;DR: The defining feature of the complexity class NP is polynomial-time verifiability: a candidate answer is cheap to check. The asymmetry that matters shows up in its hardest members, the …
- From Robot Soccer to the Agentic World Cup: Why Games Move AI Forward
🎬 Also available as a blog walkthrough video: a narrated screencast of this post. TL;DR: Twenty years ago, a year-long university project group I was part of built a team of soccer-playing cube robots …
- How My Setup Learns While You Sleep
TL;DR: The thing that makes an agent get better over time is a loop, and the loop is the same whether you buy it or build it: capture raw interaction, consolidate it on a schedule, and keep the lesson …
- Weekly Review — June 29-July 5, 2026
TL;DR: This week was about where knowledge lives, who owns it, and where the real work actually is. Three posts climb the agent-memory ladder: naming the spectrum, then the retrieval mechanism (vector …
- Your Obsidian Vault Is a Knowledge Graph
🎬 Also available as a blog walkthrough video: a narrated screencast of this post. TL;DR: Tony Seale spent six months working almost entirely inside a markdown wiki wired to an LLM, and argued the most …
- The Other Half of Two-Pizza
TL;DR: Werner Vogels’ “A return to two-pizza culture” describes how a coding agent let one scientist prototype Amazon Quick Desktop overnight, and how ownership culture carried it from there. Reading …
- Agent Memory Is a Spectrum, Not a Switch
🎬 Also available as a blog walkthrough video: a narrated screencast of this post. TL;DR: “My agent keeps forgetting things” is usually not a model problem. It’s a memory-architecture problem. Agent …
- Weekly Review — June 22-28, 2026
🎬 Also available as a blog walkthrough video — a short narrated tour of the week’s five posts. TL;DR: Five posts this week, and they rhyme. Each one treats an AI agent as a first-class user of a …
- Scrawl: Building a Tool for Humans and Agents
🎬 Also available as a blog walkthrough video with a narrated walkthrough. TL;DR: I watched someone present across half a dozen apps, drawing on screen to make their points, and I wanted the same …
- Give Every Agent Its Own Computer: AWS Lambda MicroVMs
🎬 Also available as a blog walkthrough video where I walk through the post and the demo. TL;DR: AWS Lambda MicroVMs (launched 22 June 2026) gives every user, job, or AI agent its own isolated, …
- MCP Strategies on AWS, Part 4: Governance in Code
TL;DR: The governance pillar of the AWS MCP guidance has the best whiteboard story in the whole document, so I made it runnable. The story: an admin asks an agent to clone a production database, the …
- MCP Strategies on AWS, Part 2: Tool Design in Code
TL;DR: The AWS Prescriptive Guidance paper on MCP gives the most useful, checkable rules in its tool-design section, so I wrote code to test them. A token-tax counter on a realistic 20-tool GitHub …
- From Columbo to Coworker
🎬 Also available as a blog walkthrough video with narrated diagrams. TL;DR: Running many interactive AI agents at once feels like Lieutenant Columbo lives in your terminal: every few minutes one turns …
- MCP Strategies on AWS, Part 1: Reading the Whole Guide
🎬 Also available as a blog walkthrough video with narrated diagrams. TL;DR: AWS published a Prescriptive Guidance paper, “Model Context Protocol strategies on AWS,” that organizes MCP into three …
- CLI vs MCP, Part Two: The First Gap Just Closed
TL;DR: In March I argued the CLI vs MCP debate was the wrong debate, and that the CLI’s advantages were a temporary artifact of training data, not a law of physics. One of those advantages was …
- Why Your Cheapest Model Should Write the Harness
TL;DR: A May 2026 paper separates two capabilities that self-improving agents usually conflate: writing harness updates and benefiting from them. Writing is flat across model tiers: a 9B open model …
- Architecting Skills: How Code Makes AI Agents More Reliable Over Time
TL;DR: An agent skill starts life as a markdown file full of instructions. It works, sometimes. Then you watch it fail in ways that are hard to predict, and you notice a pattern: the steps that break …
- Your Agent's Skills Are Bounded Contexts (Design Them Like It)
TL;DR: My AWS colleague Dennis Traub made the case that Domain-Driven Design’s Ubiquitous Language is now critical infrastructure for AI agents. I want to take that one layer up, into architecture. …
- AI-Assisted Talk Prep: The Recording-Analysis-Iterate Loop
TL;DR: I prepared my AWS Summit Hamburg talk through 6 recorded dry runs, each analyzed by a structured AI skill. The skill transcribes via Amazon Transcribe, counts fillers with word-level …
- Intelligence Is Collective, Not Artificial
TL;DR: Michael I. Jordan, the man Science magazine named “the world’s most influential computer scientist,” has never considered himself an AI researcher. His thesis: intelligence is not a property of …
- Your AI Judge Needs a Judge
TL;DR: Most teams I see ship LLM judges without testing them against human labels. The result: judges that are confidently, consistently wrong on 30%+ of cases. Hamel Husain’s “critique shadowing” …
- Announcements Fade. Experiences Stick. — My AWS Summit Hamburg 2026 Recap
TL;DR: AWS Summit Hamburg 2026 was two days of community, conversations, and agentic AI everywhere. My talk on production challenges landed well, the question that stuck was about getting business …
- When AI Joins the Team: Observations from a 3-Day Hack Event
Seven Teams, One Vision TL;DR: Seven teams built an AI-driven audio product in three days. AI agents didn’t replace human collaboration, they amplified individual speed, shifting the bottleneck to …
- Is the AI Subsidy Era Ending? And Why That Might Be a Good Thing
I was listening to a recent episode of The AI Daily Brief, “The AI Subsidy Era Is Over”, and my thoughts started spinning. Not because the argument was new, but because it connected dots I’d been …
- The Agent Security Stack Nobody Is Building
The Scenario Nobody Planned For It’s 11 PM. Your customer support agent, the AI one, is processing a refund request. It queries the order database, pulls the customer’s payment history, and calls the …
- The AI Track at AWS Summit Hamburg 2026: From Demo to Deployment
Last year, I wrote about the dedicated Gen AI track at AWS Summit Hamburg 2025. The response was overwhelming, the track was packed, conversations spilled into the hallways, and the Fischbrötchen at …
- MCP Sampling & Elicitation: When Servers Talk Back
From Request-Response to Collaboration When I wrote about the CLI vs MCP debate [1], I focused on the infrastructure patterns underneath. But MCP itself has been evolving, and the latest additions …
- Self-Improving Models: What MiniMax M2.7 Actually Does
The Headline vs The Reality “Model trains itself over 100+ autonomous cycles.” That was the headline when MiniMax released M2.7 on March 18, 2026 [1]. It sounds like science fiction: a model …
- Fischbrötchen and Failure Rates — I'm Speaking at AWS Summit Hamburg
Fischbrötchen and Failure Rates Last year, the AWS Summit left Berlin for Hamburg. After years of presenting at the Berlin Summit, I wasn’t sure how I’d feel about the move. Then I opened the …
- From Cloud-Native to AI-Native: What Actually Changes
The Fifteen-Year Echo Fifteen years apart. Same stage. Different world. In 2010, Adrian Cockcroft stood on the QCon stage and told the audience that Netflix was running its entire business on a public …
- The Protocol We Should Have Built for Humans
Namaste from 6,165 Meters I just summited Imja Tse (Island Peak, 6,165 meters) in Nepal. No Slack, no email, no MCP servers crashing in the background. Just ice, thin air, and the kind of clarity that …
- Security Is Job Zero — Even (Especially) in the Age of Coding Agents
$20 and Two Hours On February 28, 2026, security startup CodeWall gave an autonomous AI agent a single input: a domain name. Two hours and approximately $20 in API tokens later, the agent had full …
- AI Coding Productivity: 10%, Not 10x
The Number Nobody Wants to Hear A few weeks ago, I wrote about running my entire workday through an AI agent [1], meetings, research, CRM, content creation. Eight hours of productive work, not a …
- CLI vs MCP: The Wrong Debate
The Zombie Processes and the 50GB Cache A few weeks ago, I noticed my MacBook was sluggish. I found orphaned MCP server processes that had failed to shut down cleanly, a problem Didier Durand …
- The Coding Agent That Doesn't Code
The Friday That Wrote Itself Last Friday, I used a coding agent for eight hours straight. I didn’t write a single line of code. I prepared a customer meeting by pulling context from Slack threads, …