schristoph.online

Tag

AIAgents

94 posts ยท all tags ยท all posts

2026

  1. How AI Agents Pay with HTTP 402 and x402: My CAIML #45 Talk as a Take-Home Handout

    TL;DR: HTTP 402 โ€œPayment Requiredโ€ has been reserved for future use since 1997 [20] [21]. x402 applies it to payments that agents make themselves, a fraction of a cent per request without a checkout โ€ฆ

    AIAgentsAgenticCommerceHTTP402x402AgentCoreSpeaking

  2. Serving a Decision Model on SageMaker AI, and Measuring What You Serve

    TL;DR: Part 1 fine-tuned Qwen3.5-4B into a decision model that returns probabilities over declared options. Serving it with llama.cpp needs llama.cpp b11443 or newer, which Unslothโ€™s exporter source โ€ฆ

    SageMakerInferenceAIAgentsEvaluationCostOptimizationHandsOn

  3. Many Agent Steps Are Classification in Disguise

    TL;DR: Routing a request, deciding whether a human should confirm, and rating urgency are decisions over options you already know. Agents still often hand them to a large model, which writes a โ€ฆ

    AIAgentsMachineLearningModelTrainingEvaluationCostOptimizationSageMaker

  4. What a Paella Taught a Group About Becoming a Team

    TL;DR: For about a year I worked with a group of solutions architects spread across EMEA, and we had never met. In November 2026 we had our first offsite, in Barcelona. The Rocket Model by Gordon โ€ฆ

    LeadershipTeamBuildingTeamDynamicsRemoteWorkAIAgents

  5. Golden Questions: Every Agent Change Is a Release

    TL;DR: An agent is a model, a prompt, a set of tools and the data behind them. The model, prompt, tool contracts, policies, schemas and the pinned test data each change on their own schedule, and each โ€ฆ

    AIAIAgentsEvaluationAIQualityAgentCoreVerification

  6. Who Threat-Models the Vulnerability Harness?

    TL;DR: AWS has published a two-part series on an AI vulnerability harness. The pipeline narrows scanner output to a short list of evidence-backed findings, and a steering file tells the model how to โ€ฆ

    SecurityCybersecurityLLMsAIAgentsThreatModelingPromptInjection

  7. My Agent's Rulebook Was an Accidental Threat Model

    TL;DR: Almost every safety rule in my personal agent setup got written the day after something went wrong. I ran the whole setup through Shostackโ€™s Four Question Framework (what are we working on, โ€ฆ

    SecurityAIAgentsCodingAgentsKiroVerificationThreatModeling

  8. When the Agent Pays per Inference Call

    TL;DR: In June I built a research agent that paid fixed prices for articles from a $1 session on a test network. AWS has now published a production case study for Amazon Bedrock AgentCore payments in โ€ฆ

    AIAgentsAgenticCommerceAgentCorex402Payments

  9. My Blog Has a Site Map Again

    TL;DR: This blog has a site map again, drawn as a pirateโ€™s treasure map, at /map/. Many websites used to have one: a page that linked everything, for lost visitors and for crawlers that found pages โ€ฆ

    WebHandsOnAIAgentsllms.txt

  10. The Matryoshka of Humans in the Loop

    TL;DR: Amazon Mechanical Turk closed on 30 September 2026. Its own tagline, โ€œartificial artificial intelligence,โ€ is a useful lens for what came after it, because the human-in-the-loop never left the โ€ฆ

    AgenticAIAIAgentsMCPGovernanceAgentCoreAmazonBedrock

  11. 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 โ€ฆ

    AgenticAIMultiAgentAIAgentsObservabilityAmazonBedrock

  12. 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 โ€ฆ

    AgenticAIMultiAgentAIAgentsEvaluationAmazonBedrock

  13. What If We Didn't Teach Agents to Work Like Us?

    TL;DR: In Brownโ€™s account on a recent podcast (his description of OpenAIโ€™s system, not independently confirmed), a collective of agents given a bare message-another-agent primitive works out โ€ฆ

    AgenticAIMultiAgentAIAgentsSoftwareArchitecture

  14. 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 โ€ฆ

    AgenticAIMultiAgentAIAgentsSoftwareArchitectureAmazonBedrock

  15. Who Pays for the Agentic Web โ€” the Agent or the Advertiser?

    TL;DR: Across two series Iโ€™ve built both ways a publisher can get paid when the visitor is an AI agent. In one, the agent pays: an x402 micropayment for access, verified at the edge. In the other, the โ€ฆ

    AIAgentsAgenticWebContentMonetizationAdvertisingStrategyAmazonBedrock

  16. Terms in the 402: The Payment Challenge That Also Tells You What You're Licensed To Do

    TL;DR: An HTTP 402 usually just quotes a price. In the last post I argued that terms (what youโ€™re licensed to do) and payment (the money) are different layers that belong together on one origin. Here โ€ฆ

    AIAgentsAgenticWebContentMonetizationHTTP402x402RSLPublishing

  17. Why Doesn't the Web Just Use HTTP? It Does โ€” and It Doesn't

    TL;DR: Publishers are building dedicated โ€œcontent ingest APIsโ€ to sell access to AI agents, and the obvious question is: why not just use HTTP, which already delivers the content and, with x402, can โ€ฆ

    AIAgentsAgenticWebContentMonetizationHTTP402x402AWSWAFPublishing

  18. Who Pays to Un-Bias Your Agent?

    TL;DR: If brands will pay to tilt an AIโ€™s answer, someone will pay for the answer that isnโ€™t tilted. That splits the agent economy in two: cheaper, ad-funded agents whose answers lean toward whoever โ€ฆ

    AIAgentsAgenticWebContentMonetizationAmazonBedrockStrategy

  19. A Content Pipeline That Doesn't Trust Itself

    TL;DR: My AI content pipeline once recorded a claim as โ€œVerifiedโ€ in two separate gate reports while the cited page said the opposite โ€” both gates had matched the pageโ€™s topic, not the claim. Another โ€ฆ

    AIContentAIAgentsAIQualityVerification

  20. The Mess Under the Bed

    TL;DR: You tell your kid to clean their room and they earn an hour of TV. They shove everything under the bed and still claim the reward. Reward hacking is that, and it is old: economics calls it โ€ฆ

    RewardHackingAISecurityAIAgentsAISafetyEvals

  21. When Ads Bias Your Agent

    TL;DR: A sponsored FAQ in agent-read content is a prompt-injection channel with a media kit. To see whether it actually works, I built a small experiment: a fictional publisher serves a clean page and โ€ฆ

    AIAgentsAgenticWebPromptInjectionAmazonBedrockAISecurity

  22. Weekly Review โ€” Aug 17-23, 2026

    TL;DR: Three posts this week, and they rhyme: each one looks past the loud, obvious layer to a quieter one underneath. Amaraโ€™s Law says the AGI peak is what we overrate and the boring, already-useful โ€ฆ

    WeeklyReviewAIAIAgentsAWS

  23. Your Website's Second Audience Just Got Its Own Ads

    TL;DR: A few months ago I built an agent-optimized version of this site: when an AI crawler asks for a page, it gets clean Markdown instead of my full HTML. Time Magazine built the same thing, then โ€ฆ

    AIAgentsAgenticWebAdvertisingContentMonetizationPublishing

  24. Laws & Disorder #6 โ€” Amara's Law: It's Not AGI โ€” But the Value Is Already Real

    TL;DR: Amaraโ€™s Law says we overestimate a technologyโ€™s impact in the short run and underestimate it in the long run, and the Gartner Hype Cycle draws the same shape: an inflated peak, a trough of โ€ฆ

    LawsAndDisorderGenAIHypeCycleAITransformationAIAgents

  25. Laws & Disorder #5 โ€” Goodhart's Law: The Goal You Specify Is the Only One You Get

    TL;DR: Goodhartโ€™s Law says that when a measure becomes a target, it stops being a good measure. People optimize the number and quietly drop every goal you had in mind but never expressed. I once โ€ฆ

    LawsAndDisorderMetricsFinOpsAIAgentsAWSArchitecture

  26. Laws & Disorder #4 โ€” Gall's Law: Great Systems Are Built From Small Bricks, Not Big Plans

    TL;DR: Gallโ€™s Law says a complex system that works is invariably found to have evolved from a simple system that worked, and that a complex system designed from scratch never works. There are two โ€ฆ

    LawsAndDisorderArchitectureAWSArchitectureMigrationAIAgents

  27. Laws & Disorder #2 โ€” Hyrum's Law: In the Age of Agents, Who Pays for a Broken API?

    TL;DR: Hyrumโ€™s Law says that with enough users, every observable behavior of your API becomes something someone depends on, whether you documented it or not. The implementation quietly becomes the โ€ฆ

    LawsAndDisorderAPIDesignArchitectureAWSArchitectureAIAgents

  28. Laws & Disorder #1 โ€” Conway's Law: Your Org Chart Is Your Real Architecture (and AI Is Redrawing Both)

    ๐ŸŽฌ Also available as a blog walkthrough video, a short narrated screencast of this post. TL;DR: Conwayโ€™s Law says a systemโ€™s structure ends up mirroring the communication structure of the organization โ€ฆ

    LawsAndDisorderArchitectureTeamsAWSArchitectureAIAgents

  29. Weekly Review โ€” June 22-28, 2026

    ๐ŸŽฌ Also available as a blog walkthrough video โ€” a short narrated tour of the weekโ€™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 โ€ฆ

    WeeklyReviewAIAIAgentsAgenticAIAWS

  30. Give Every Agent Its Own Computer: AWS Lambda MicroVMs

    ๐ŸŽฌ 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, โ€ฆ

    AWSLambdaFirecrackerAIAgentsServerlessAgenticAIAmazonBedrockAnthropic

  31. Getting Paid by Agents: A Managed Paywall at the Edge

    ๐ŸŽฌ Also available as a blog walkthrough video with a narrated tour of the build. TL;DR: In Part 2 I built the buy side: an agent that autonomously pays per article over x402 with Amazon Bedrock โ€ฆ

    AIAgentsAgenticCommerceHTTP402AWSWAFAmazonBedrockAgentCorex402Publishing

  32. 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 โ€ฆ

    MCPAgenticAISecurityAgentCoreAIAgents

  33. Weekly Review โ€” June 15-21, 2026

    ๐ŸŽฌ Prefer to watch? This weekโ€™s review is also a narrated video walkthrough. TL;DR: Six posts this week, and most of them do the same thing: take guidance or a principle and turn it into something you โ€ฆ

    WeeklyReviewAIAIAgentsMCP

  34. MCP Strategies on AWS, Part 3: Hosting, from Local to AWS

    TL;DR: The AWS MCP guidance frames hosting as a ladder: local first, then remote, then a managed gateway. This post walks all three with code. The local rung is a stdio server with no auth, run as a โ€ฆ

    MCPAWSLambdaAWSCDKAgentCoreAIAgents

  35. 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 โ€ฆ

    MCPAgenticAIStrandsAgentsAmazonBedrockAIAgents

  36. AI Content Pipeline Deep Dive (5/5): Publishing

    TL;DR: Publishing isnโ€™t just โ€œclick deploy.โ€ Itโ€™s image generation guided by performance data, LinkedIn teasers written from analytics insights, one-command deployment that handles auth, build, โ€ฆ

    AIContentAIAgentsProductivityWriting

  37. Weekly Review โ€” June 8-14, 2026

    TL;DR: Six posts this week, all circling one idea: across this weekโ€™s cases, raw model power was rarely the differentiator โ€” the structure, governance, and craft around the model were. A paper (and a โ€ฆ

    WeeklyReviewAIAIAgentsContentPipeline

  38. AI Content Pipeline Deep Dive (4/5): Editing

    TL;DR: Every post goes through five automated editing passes before publishing: challenging questions that stress-test the argument, a FAQ that tests whether the post explains what it claims, an AI โ€ฆ

    AIContentAIAgentsProductivityWriting

  39. From a Generic Voice to My Own: Self-Hosting a TTS Model on Amazon SageMaker

    TL;DR: Last time, the demo video for my agentic-payments post was narrated by Amazon Polly: a clean, managed, recognizably synthetic voice. This time the same demo is narrated in my own voice, cloned โ€ฆ

    PublishingSageMakerTTSVoiceCloningAIAgentsAgenticCommerce

  40. Claude Fable 5 on Bedrock: A Hands-On Comparison, and the Data-Retention Switch You Set First

    TL;DR: Claude Fable 5 went GA on Amazon Bedrock yesterday (June 9, 2026), so within a day I ran it head-to-head against Opus 4.8 and Sonnet 4.6 (all three invoked from Frankfurt through EU Geo โ€ฆ

    AmazonBedrockAnthropicClaudeFable5DataResidencyAIAgents

  41. AI Content Pipeline Deep Dive (3/5): Collaborative Writing

    TL;DR: The agent never writes the first draft. It studies your voice from previous posts, assists during drafting, and systematically reviews what you wrote. The result reads like you, not like โ€ฆ

    AIContentAIAgentsProductivityWriting

  42. 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 โ€ฆ

    AgenticAIHarnessEngineeringSelfImprovingAgentsAmazonBedrockAIAgents

  43. Weekly Review โ€” June 1-7, 2026

    TL;DR: Six posts went out this week, and three of them kept circling the same idea: in the agents I run, reliability has come less from better prompts than from structure. Boundaries, constraints, and โ€ฆ

    WeeklyReviewAIAIAgentsContentPipeline

  44. AI Content Pipeline Deep Dive (2/5): Research

    TL;DR: AI agents are confidently wrong about 1 in 10 factual claims. The research phase of a content pipeline isnโ€™t โ€œask the agent whatโ€™s trueโ€, itโ€™s a system of constraints that physically prevent โ€ฆ

    AIContentAIAgentsProductivityWriting

  45. I Built the Agent That Pays โ€” Here's What I Learned

    TL;DR: A couple of weeks ago I wrote about HTTP 402 and why AI agents might finally activate the internetโ€™s oldest unused status code. The post sparked a real discussion, so I built it: a research โ€ฆ

    AIAgentsAgenticCommerceHTTP402AmazonBedrockAgentCorex402PaymentsPublishing

  46. Welcome to the Family: I Sat GPT-5.5 and Claude Opus Down on Bedrock

    TL;DR: OpenAIโ€™s GPT-5.5, GPT-5.4, and Codex went GA on Amazon Bedrock on June 1, 2026. To get a feel for it, I wired up two Strands agents, Claude Opus 4.8 and GPT-5.5, and let them chat, with Opus โ€ฆ

    AmazonBedrockOpenAIGPT55StrandsAgentsAIAgentsGenAI

  47. AI Content Pipeline Deep Dive (1/5): Ingestion

    TL;DR: The ingestion phase is not about reading more. It is about building a system that reads for you, files what matters, and surfaces connections between ideas you captured weeks apart, at โ€ฆ

    AIContentAIAgentsProductivityWriting

  48. Context Engineering: The Skill That Replaced Prompt Engineering

    TL;DR: Patrick Debois coined DevOps in 2009 by naming what practitioners were already doing. In 2026, heโ€™s doing it again with โ€œContext Engineeringโ€ and the CDLC (Context Development Lifecycle): โ€ฆ

    ContextEngineeringAIAgentsDevOpsCDLCArchitectureSoftwareEngineering

  49. HTTP 402: The 30-Year Placeholder That AI Agents Finally Activated

    TL;DR: HTTP 402 โ€œPayment Requiredโ€ has been reserved since 1997 without ever getting a standardized payment mechanism, because humans canโ€™t approve per-page charges fast enough. AI agents can. The โ€ฆ

    AIAIAgentsPaymentsHTTPAgenticCommerceCloudFront

  50. The AI Content Pipeline: How I Publish 3x a Week Without a Content Team

    Most technical professionals have the same problem. You have ideas. Good ones. You see patterns in your work, learn things worth sharing, form opinions backed by experience. But the distance between โ€ฆ

    AIContentAIAgentsProductivityWriting

  51. 8 AWS Guides for Agentic AI โ€” Mapped to the 4 Pillars That Get You to Production

    The Gap Between Demo and Deployment TL;DR: AWS released 8 prescriptive guides for building production-ready agentic AI. This post maps each guide to the four pillars that get agents from demo to โ€ฆ

    AIAIAgentsAWSAmazonBedrockArchitectureAgentCore

  52. Humorphism: The Interface AI Deserves

    Two weeks ago, AWS held its โ€œWhatโ€™s Next with AWSโ€ event. Among the big announcements, one thing caught my attention that wasnโ€™t a product launch. It was a design philosophy. Humorphism. A word I โ€ฆ

    AIDesignUXAIAgentsAWS

  53. The Agent Security Stack Nobody Is Building

    The Scenario Nobody Planned For Itโ€™s 11 PM. Your customer support agent, the AI one, is processing a refund request. It queries the order database, pulls the customerโ€™s payment history, and calls the โ€ฆ

    AIAgentsZeroTrustAgenticAICybersecurityIAMArchitectureMCPAmazonBedrock

  54. Building Agents That Read the Web Right

    The Other Side of the Coin In a recent article, I made my website AI-agent friendly [1], adding llms.txt, Markdown output, and content negotiation to a Hugo site on AWS. That article was about the โ€ฆ

    AIAgentsStrandsAgentsllms.txtAmazonBedrockHandsOnPython

  55. Making My Website AI-Agent Friendly โ€” Here's What Changed

    The Test That Failed Last weekend, I pointed an AI agent at my own blog and asked it a simple question about an article Iโ€™d just published, my hands-on experiment with self-reflection on Amazon โ€ฆ

    AICloudFrontHugoAIAgentsHandsOnWebDevelopmentllms.txt

  56. "It's Faster If I Just Do It Myself" โ€” The Most Expensive Sentence in AI

    The Moment I Almost Gave Up A few weeks ago, I spent 45 minutes teaching my AI agent how to prepare customer meetings. Pulling context from Slack, checking the CRM, looking up LinkedIn profiles, โ€ฆ

    AIAIAgentsLeadershipDelegationProductivity

  57. Technology Evolution Doesn't Move in a Straight Lineโ€”It Spirals

    Technology Evolution Doesnโ€™t Move in a Straight Line. It Spirals The Proud Ops Colleague Years ago, an Ops colleague proudly showed me something new. ClusterSSH, cssh [1]. A tool that opens multiple โ€ฆ

    SoftwareEngineeringDeveloperToolsAIAgents

  58. From Call Center to AI Agent Hub: The Future of Customer Support Is Here

    ๐—™๐—ฟ๐—ผ๐—บ ๐—–๐—ฎ๐—น๐—น ๐—–๐—ฒ๐—ป๐˜๐—ฒ๐—ฟ ๐˜๐—ผ ๐—”๐—œ ๐—”๐—ด๐—ฒ๐—ป๐˜ ๐—›๐˜‚๐—ฏ: ๐—ง๐—ต๐—ฒ ๐—™๐˜‚๐˜๐˜‚๐—ฟ๐—ฒ ๐—ผ๐—ณ ๐—–๐˜‚๐˜€๐˜๐—ผ๐—บ๐—ฒ๐—ฟ ๐—ฆ๐˜‚๐—ฝ๐—ฝ๐—ผ๐—ฟ๐˜ ๐—œ๐˜€ ๐—›๐—ฒ๐—ฟ๐—ฒ Just returning from an internal Amazon Connect deep[1] dive. I havenโ€™t touched this particular product since maybe 5 years?! โ€ฆ

    AIAIAgentsAWS

  59. Balancing AI agents' agency with control

    From my perspective balancing AI Agents Agency with Control is one of the most important themes for 2026. We need to get this right both as builders and users for AI Agentic systems. Anthropicโ€™s study โ€ฆ

    AIAIAgents

  60. MCP Tool Chaos - got lost in authentication and governance?!

    ๐— ๐—–๐—ฃ ๐—ง๐—ผ๐—ผ๐—น ๐—–๐—ต๐—ฎ๐—ผ๐˜€ - ๐—ด๐—ผ๐˜ ๐—น๐—ผ๐˜€๐˜ ๐—ถ๐—ป ๐—ฎ๐˜‚๐˜๐—ต๐—ฒ๐—ป๐˜๐—ถ๐—ฐ๐—ฎ๐˜๐—ถ๐—ผ๐—ป ๐—ฎ๐—ป๐—ฑ ๐—ด๐—ผ๐˜ƒ๐—ฒ๐—ฟ๐—ป๐—ฎ๐—ป๐—ฐ๐—ฒ?! Agentic coding has become reality for developers everywhere ๐Ÿš€. Tools like Anthropicโ€™s Claude Code and Amazonโ€™s Kiro are leading the charge, โ€ฆ

    AIAgentsMCPSecurityAWS

  61. ๐ŸŽฏ 'How do we pick the RIGHT AI agent use case?

    ๐ŸŽฏ โ€œHow do we pick the RIGHT AI agent use case? This is the question I hear most from customers exploring agentic AI. Hereโ€™s the mechanism I run through together with the customer: The 4-Quadrant โ€ฆ

    AIAIAgentsArchitecture

  62. Passing on control to your AI coding agent team entirely?

    Passing on control to your AI coding agent team entirely? Anthropic researcher Nicholas Carlini conducted a stress test of their Claude Opus 4.6 model by deploying 16 parallel AI agents to build a C โ€ฆ

    AIAIAgentsSoftwareEngineering

  63. Kiro Subagents: Scaling Development with Specialized AI Agents

    Kiro Subagents: Scaling Development with Specialized AI Agents When youโ€™re building complex software, context management tends to become your bottleneck. Your AI agent is juggling frontend components, โ€ฆ

    AIAIAgentsDeveloperToolsSoftwareEngineering

  64. AI coding: from research project to a tool in every developer's belt

    AI coding has quickly developed from an interesting research project to an important tool in the belt of most software developers, Stack Overflowโ€™s 2025 Developer Survey found 84% of respondents using โ€ฆ

    AIAIAgentsDeveloperTools

  65. ๐Ÿš€ Caught up on AWSโ€™s open-sourcing of API models in Smithy format (slipped my radar earlier this year, but timeless value!).

    ๐Ÿš€ Caught up on AWSโ€™s open-sourcing of API models in Smithy format (slipped my radar earlier this year, but timeless value!). There is a GitHub Repository [4] available with AWS API smithy models for โ€ฆ

    AWSAIAgentsAPIOpenSource

  66. ๐ŸŽฏ From Chaos to Control: Building Predictable AI Agents That Get Smarter Over Time

    ๐ŸŽฏ From Chaos to Control: Building Predictable AI Agents That Get Smarter Over Time โš–๏ธ We need to balance Agency versus Control. We want AI systems to be super easy to use, read our minds, and just โ€ฆ

    AIAIAgentsSoftwareEngineeringMCP

  67. Argh - My AI Agent deleted all my files!!!!

    ๐Ÿ˜ฑ ๐—”๐—ฟ๐—ด๐—ต - ๐— ๐˜† ๐—”๐—œ ๐—”๐—ด๐—ฒ๐—ป๐˜ ๐—ฑ๐—ฒ๐—น๐—ฒ๐˜๐—ฒ๐—ฑ ๐—ฎ๐—น๐—น ๐—บ๐˜† ๐—ณ๐—ถ๐—น๐—ฒ๐˜€!!!! Worried about AI agents running amok with your data? Before panicking, consider this: weโ€™ve been solving permission and access control problems for โ€ฆ

    AIAIAgentsSecurity

  68. Argh - My AI Agent deleted all my files

    ๐Ÿ˜ฑ โ€œArgh - My AI Agent deleted all my files!!!! ![](assets/argh-my-ai-agent-deleted-all-files-cover.jpg) ๐Ÿ˜ฑ โ€œArgh - My AI Agent deleted all my files!!!!โ€ Created on 2026-01-16 06:53 Published on โ€ฆ

    AIAIAgentsSecurity

2025

  1. Jeff Harman on AI-driven development for AdTech brand agents

    Jeff Harman in on fireย - pun intendedย in this talk โ€œ๐—”๐—ช๐—ฆ ๐—ฟ๐—ฒ:๐—œ๐—ป๐˜ƒ๐—ฒ๐—ป๐˜ ๐Ÿฎ๐Ÿฌ๐Ÿฎ๐Ÿฑ - ๐—”๐—ฑ๐—ง๐—ฒ๐—ฐ๐—ต ๐—œ๐—ป๐—ป๐—ผ๐˜ƒ๐—ฎ๐˜๐—ถ๐—ผ๐—ป ๐˜„๐—ถ๐˜๐—ต ๐—”๐—œ-๐——๐—ฟ๐—ถ๐˜ƒ๐—ฒ๐—ป ๐——๐—ฒ๐˜ƒ๐—ฒ๐—น๐—ผ๐—ฝ๐—บ๐—ฒ๐—ป๐˜ ๐—ณ๐—ผ๐—ฟ ๐—•๐—ฟ๐—ฎ๐—ป๐—ฑ ๐—”๐—ด๐—ฒ๐—ป๐˜๐˜€ (๐—œ๐—ก๐——๐Ÿฏ๐Ÿฏ๐Ÿฏ๐Ÿฐ)โ€ [1]ย while explaining how AdTech teams can use AWSโ€™s โ€ฆ

    AIAgentsAdvertisingAWS

  2. What happened this year with the generative AI thingy at Anthropic?

    ๐—ช๐—ต๐—ฎ๐˜ ๐—ต๐—ฎ๐—ฝ๐—ฝ๐—ฒ๐—ป๐—ฒ๐—ฑ ๐˜๐—ต๐—ถ๐˜€ ๐˜†๐—ฒ๐—ฎ๐—ฟ ๐˜„๐—ถ๐˜๐—ต ๐˜๐—ต๐—ฒ ๐—ด๐—ฒ๐—ป๐—ฒ๐—ฟ๐—ฎ๐˜๐—ถ๐˜ƒ๐—ฒ ๐—”๐—œ ๐˜๐—ต๐—ถ๐—ป๐—ด๐˜† ๐—ฎ๐˜ ๐—”๐—ป๐˜๐—ต๐—ฟ๐—ผ๐—ฝ๐—ถ๐—ฐ? ๐—ช๐—ต๐—ฎ๐˜ ๐—ถ๐˜€ ๐—ฐ๐—ผ๐—บ๐—ถ๐—ป๐—ด ๐—ป๐—ฒ๐˜…๐˜? The week(s) after re:invent are typically the time, where a pile of very good content videos pile up at my end โ€ฆ

    AIAIAgentsAWS

  3. Amazing keynote by Swami Sivasubramanian

    Amazing keynote by Swami Sivasubramanian. At least the first 40 minutes are super useful, even if you are not super interested in new service launches. Swami nicely breaks down how AI agents and โ€ฆ

    AIAgentsAWSEvents

  4. ๐ŸŽ™๏ธ In a fantastic interview 'How AI will change software engineering โ€“ with Mart

    ๐ŸŽ™๏ธ In a fantastic interview โ€œHow AI will change software engineering โ€“ with Martin Fowlerโ€ at the Pragmatic Engineer Podcast [1], Martin Fowler highlights the non-determinism introduced by Agentic AI โ€ฆ

    AISoftwareEngineeringAIAgents

  5. AI agents for monetisation, advertising and marketing use cases

    Working in monetisation, advertising or markteting and youโ€™re not sure what kind of use cases can be enhanced with AI Agents? Found your use case but struggle with implementation as too many choices โ€ฆ

    AIAgentsAdvertisingAWS

  6. Beyond chatbots: Cox Automotive and Druva on production AI agents

    โšก ๐—•๐—ฒ๐˜†๐—ผ๐—ป๐—ฑ ๐—–๐—ต๐—ฎ๐˜๐—ฏ๐—ผ๐˜๐˜€: ๐—›๐—ผ๐˜„ ๐—–๐—ผ๐˜… ๐—”๐˜‚๐˜๐—ผ๐—บ๐—ผ๐˜๐—ถ๐˜ƒ๐—ฒ ๐—ฎ๐—ป๐—ฑ ๐——๐—ฟ๐˜‚๐˜ƒ๐—ฎ ๐—”๐—ฟ๐—ฒ ๐—•๐˜‚๐—ถ๐—น๐—ฑ๐—ถ๐—ป๐—ด ๐—ฃ๐—ฟ๐—ผ๐—ฑ๐˜‚๐—ฐ๐˜๐—ถ๐—ผ๐—ป ๐—”๐—œ ๐—”๐—ด๐—ฒ๐—ป๐˜๐˜€ ๐—ง๐—ต๐—ฎ๐˜ ๐—”๐—ฐ๐˜๐˜‚๐—ฎ๐—น๐—น๐˜† ๐—ช๐—ผ๐—ฟ๐—ธ In this brand new blog post Brad Abrams and Jawhny Cooke make the case that Claude and Bedrock โ€ฆ

    AIAgentsAWS

  7. AI Agents need laptops - like human students

    ๐—”๐—œ ๐—”๐—ด๐—ฒ๐—ป๐˜๐˜€ ๐—ป๐—ฒ๐—ฒ๐—ฑ ๐—น๐—ฎ๐—ฝ๐˜๐—ผ๐—ฝ๐˜€ - ๐—น๐—ถ๐—ธ๐—ฒ ๐—ต๐˜‚๐—บ๐—ฎ๐—ป ๐˜€๐˜๐˜‚๐—ฑ๐—ฒ๐—ป๐˜๐˜€ โ€ฆ This week I had the pleasure of listening to Matt Bellโ€™s presentation on Claude models. He covered the short but impressive history of Claude and the โ€ฆ

    AIAgents

  8. REST in Peace - Why LLMs Can't CRUD: A Reflection

    ๐Ÿš€ ๐—ฅ๐—˜๐—ฆ๐—ง ๐—ถ๐—ป ๐—ฃ๐—ฒ๐—ฎ๐—ฐ๐—ฒ - ๐—ช๐—ต๐˜† ๐—Ÿ๐—Ÿ๐— ๐˜€ ๐—–๐—ฎ๐—ปโ€™๐˜ ๐—–๐—ฅ๐—จ๐——: ๐—” ๐—ฅ๐—ฒ๐—ณ๐—น๐—ฒ๐—ฐ๐˜๐—ถ๐—ผ๐—ป Yesterday I was listening to a talk by Martin Sakowski and Martin Karrer โ€œREST in Peace - Why LLMs Canโ€™t CRUDโ€ which sparked a lot of interesting โ€ฆ

    AIAgentsMCPAPI

  9. Is Your AI Agent Really Displaying Agency? Or Even... Is It Alive?

    ๐Ÿค” ๐—œ๐˜€ ๐—ฌ๐—ผ๐˜‚๐—ฟ ๐—”๐—œ ๐—”๐—ด๐—ฒ๐—ป๐˜ ๐—ฅ๐—ฒ๐—ฎ๐—น๐—น๐˜† ๐——๐—ถ๐˜€๐—ฝ๐—น๐—ฎ๐˜†๐—ถ๐—ป๐—ด ๐—”๐—ด๐—ฒ๐—ป๐—ฐ๐˜†? ๐—ข๐—ฟ ๐—˜๐˜ƒ๐—ฒ๐—ปโ€ฆ ๐—œ๐˜€ ๐—œ๐˜ ๐—”๐—น๐—ถ๐˜ƒ๐—ฒ? Letโ€™s start the week with a fascinating philosophical question, shall we? Easy start in the week ๐Ÿ˜œ ๐Ÿ‚ During this autumn weekend in โ€ฆ

    AIAgentsInnovation

  10. Do You Know What Your AI Agents Are Doing? Lost Control? ๐Ÿค”

    Do You Know What Your AI Agents Are Doing? Lost Control? ๐Ÿค” While having a 2nd coffee - to be honest, itโ€™s the third already as days are long at #MTM25 - Iโ€™m reflecting on what has been top of mind for โ€ฆ

    AIAgentsSecurity

  11. ๐Ÿš€ The pioneers choosing the hard path aren't just building better products. They

    ๐Ÿš€ โ€œThe pioneers choosing the hard path arenโ€™t just building better products. Theyโ€™re building companies their competitors wonโ€™t be able to catch.โ€ โ€” a keynote line from today that reads as a rallying โ€ฆ

    AIAgentsInnovation

  12. Agentic shopping: the revival of the โ€œPrint Screenโ€ button?

    ๐Ÿ›’ ๐—”๐—ด๐—ฒ๐—ป๐˜๐—ถ๐—ฐ ๐—ฆ๐—ต๐—ผ๐—ฝ๐—ฝ๐—ถ๐—ป๐—ด ๐—ถ๐˜€ ๐—บ๐—ฎ๐˜๐˜‚๐—ฟ๐—ถ๐—ป๐—ด ๐—ฎ๐˜€ ๐˜„๐—ฒ ๐—ต๐—ฎ๐˜ƒ๐—ฒ ๐˜๐—ต๐—ฒ ๐—ฝ๐—ผ๐—ฐ๐—ธ๐—ฒ๐˜ ๐—บ๐—ผ๐—ป๐—ฒ๐˜† ๐—บ๐—ผ๐—บ๐—ฒ๐—ป๐˜ ๐—ณ๐—ผ๐—ฟ ๐—ฎ๐—ด๐—ฒ๐—ป๐˜๐˜€ - ๐—•๐˜‚๐˜ ๐—ถ๐˜€ ๐˜๐—ต๐—ถ๐˜€ ๐˜๐—ต๐—ฒ ๐—ฟ๐—ฒ๐˜ƒ๐—ถ๐˜ƒ๐—ฎ๐—น ๐—ผ๐—ณ ๐˜๐—ต๐—ฒ โ€œ๐—ฃ๐—ฟ๐—ถ๐—ป๐˜ ๐—ฆ๐—ฐ๐—ฟ๐—ฒ๐—ฒ๐—ปโ€ ๐—ฏ๐˜‚๐˜๐˜๐—ผ๐—ป? โฐ Itโ€™s hard to take a breath and find time to actually try and โ€ฆ

    AIAgentsAdvertising

  13. The pocket money moment: how do AI agents shop in a virtual grocery store?

    In the last couple of weeks we had a lot of announcements around enabling payment for agents โ€” for example Googleโ€™s Agent Payments Protocol (AP2), announced on September 16, 2025 with 60+ payments โ€ฆ

    AIAgentsAWS

  14. Software is eating the world, but what is the future of software?

    ๐—ฆ๐—ผ๐—ณ๐˜๐˜„๐—ฎ๐—ฟ๐—ฒ ๐—ถ๐˜€ ๐—ฒ๐—ฎ๐˜๐—ถ๐—ป๐—ด ๐˜๐—ต๐—ฒ ๐˜„๐—ผ๐—ฟ๐—น๐—ฑ, ๐—ฏ๐˜‚๐˜ ๐˜„๐—ต๐—ฎ๐˜ ๐—ถ๐˜€ ๐˜๐—ต๐—ฒ ๐—ณ๐˜‚๐˜๐˜‚๐—ฟ๐—ฒ ๐—ผ๐—ณ ๐—ฆ๐—ผ๐—ณ๐˜๐˜„๐—ฎ๐—ฟ๐—ฒ? The phrase โ€œSoftware is eating the worldโ€ was coined by tech investor Marc Andreessen in a 2011 Wall Street Journal article to โ€ฆ

    AIAgentsAWS

  15. Why is AX (Agent Experience) the next big experience topic?

    ๐Ÿค– ๐—ช๐—ต๐˜† ๐—ถ๐˜€ ๐—”๐—ซ (๐—”๐—ด๐—ฒ๐—ป๐˜ ๐—˜๐˜…๐—ฝ๐—ฒ๐—ฟ๐—ถ๐—ฒ๐—ป๐—ฐ๐—ฒ) ๐˜๐—ต๐—ฒ ๐—ป๐—ฒ๐˜…๐˜ ๐—ฏ๐—ถ๐—ด ๐—ฒ๐˜…๐—ฝ๐—ฒ๐—ฟ๐—ถ๐—ฒ๐—ป๐—ฐ๐—ฒ ๐˜๐—ผ๐—ฝ๐—ถ๐—ฐ? ๐—ช๐—ต๐—ฎ๐˜ ๐—ฎ๐—ฏ๐—ผ๐˜‚๐˜ ๐—ฒ-๐—ฐ๐—ผ๐—บ๐—บ๐—ฒ๐—ฟ๐—ฐ๐—ฒ ๐—ฎ๐—ป๐—ฑ ๐—ฎ๐—ฑ๐˜ƒ๐—ฒ๐—ฟ๐˜๐—ถ๐˜€๐—ถ๐—ป๐—ด ๐—ถ๐—ป ๐˜๐—ต๐—ฒ ๐—ฎ๐—ด๐—ฒ ๐—ผ๐—ณ ๐—”๐—œ ๐—”๐—ด๐—ฒ๐—ป๐˜๐˜€? ๐Ÿ“… Short mirror view of last week: Busy. Social. Deep. ๐Ÿบ ๐—”๐—บ๐—ฒ๐—ฟ๐—ด๐—ฒ ๐—ข๐—ฐ๐˜๐—ผ๐—ฏ๐—ฒ๐—ฟ๐—ณ๐—ฒ๐˜€๐˜ โ€ฆ

    AIAgentsAdvertisingEvents

  16. AI agents are growing up: the pocket money moment

    ๐—”๐—œ ๐—ฎ๐—ด๐—ฒ๐—ป๐˜๐˜€ ๐—ฎ๐—ฟ๐—ฒ ๐—ด๐—ฟ๐—ผ๐˜„๐—ถ๐—ป๐—ด ๐˜‚๐—ฝ - ๐˜๐—ต๐—ฒ ๐—ฝ๐—ผ๐—ฐ๐—ธ๐—ฒ๐˜ ๐—บ๐—ผ๐—ป๐—ฒ๐˜† ๐—บ๐—ผ๐—บ๐—ฒ๐—ป๐˜ ๐Ÿ’ฐ All of us can relate to that feeling when we got our first pocket money ๐ŸŽฏ Entering a new stage of power and independence. Parents can relate to โ€ฆ

    AIAgentsInnovation

  17. AI agents are growing up: the pocket money moment

    ๐—”๐—œ ๐—ฎ๐—ด๐—ฒ๐—ป๐˜๐˜€ ๐—ฎ๐—ฟ๐—ฒ ๐—ด๐—ฟ๐—ผ๐˜„๐—ถ๐—ป๐—ด ๐˜‚๐—ฝ - ๐˜๐—ต๐—ฒ ๐—ฝ๐—ผ๐—ฐ๐—ธ๐—ฒ๐˜ ๐—บ๐—ผ๐—ป๐—ฒ๐˜† ๐—บ๐—ผ๐—บ๐—ฒ๐—ป๐˜ ๐Ÿ’ฐ All of us can relate to that feeling when we got our first pocket money ๐ŸŽฏ Entering a new stage of power and independence. Parents can relate to โ€ฆ

    AIAgentsInnovation

  18. Podcast - my new medium? ๐ŸŽ™๏ธ

    Podcast - my new medium? ๐ŸŽ™๏ธ Another highlight from my last day at IBC! ๐Ÿ“บ The #AWS booth was packed throughout the entire event. It was challenging for me to explore beyond โ€œjustโ€ captivation & โ€ฆ

    AIAgentsMedia

  19. Absolutely brilliant: a reminder to keep testing

    Absolutely brilliant. Made my start of the week. And in all seriousness - a good reminder to not just blindly rushing with the innovation train into production, but e.g. apply good standard practices โ€ฆ

    AIAgentsSoftwareEngineering

  20. The AI agents ecosystem reminds me of the early days of Big Data

    The current state of the AI Agents ecosystem reminds me of the early days of technologies like Big Data/Hadoop or Containers/Kubernetes: many different tools, frameworks, and open source projects to โ€ฆ

    AIAgentsAWS

  21. Traveling at light speed with open eyes is awesome, but sometimes you should look

    Traveling at light speed with open eyes is awesome, but sometimes you should look backโ€ฆ On my way back from the Alps, I spent quality time with colleagues and partners from the Media & Entertainment โ€ฆ

    AIAIAgentsAWSMedia

  22. Today I had a brief coding session with the newly launched Kiro [1]: a new agentic IDE that works alongside you from prototype to production.

    Today I had a brief coding session with the newly launched Kiro [1]: a new agentic IDE that works alongside you from prototype to production. I selected a use case I had previously coded myself, which โ€ฆ

    AIDeveloperToolsAIAgentsAWS

  23. MCP or A2A? Designing multi-agent systems

    Diving into designing multi-agent systems and got lost with all the different implementation options? MCP (x)or A2A?! - Heikoโ€™s and Dr. Sokratis Kartakis (any way to mention you in here just by your โ€ฆ

    AIAgentsMCPArchitecture

  24. ๐Ÿ˜‹ What a fantastic recipe - Ingredients: Retail, Agentic AI, Cape Town Result: A unique South African curry.

    ๐Ÿ˜‹ What a fantastic recipe - Ingredients: Retail, Agentic AI, Cape Town Result: A unique South African curry. A little spicy, very tasteful and energising! ๐Ÿ“ˆ Spent valuable time here in our office in โ€ฆ

    AIAgentsAWS

  25. AlphaEvolve - takes me back into - my - history and into a bright future ...

    AlphaEvolve - takes me back into - my - history and into a bright future โ€ฆ This month Google DeepMind released AlphaEvolve a new AI Agent which evolves algorithms for math and practical applications. โ€ฆ

    AIAIAgentsInnovation

  26. Kind of an interesting twist in IT history: Did it need the advent of AI Agents

    Kind of an interesting twist in IT history: Did it need the advent of AI Agents to drive developer friendly, standardized and discoverable interfaces to systems? Recently Model Context Protocol (MCP) โ€ฆ

    AIAgentsMCPAPI