Field notes on AI that does real work.
Concrete writing on AI agents in Slack and Microsoft Teams, automating the tools your team already runs - Notion, Linear, Jira, Zendesk, GitHub - and what’s genuinely new in MCP, open models, and agents. No hype, no filler.
1,000 free credits to start. No credit card. Working in five minutes.
Latest field note
Does Microsoft Agent Framework Replace AutoGen and Semantic Kernel?
Microsoft Agent Framework 1.0 shipped April 3, 2026, folding AutoGen and Semantic Kernel into one SDK. Here is what actually changed, what the migration costs, and what the marketing elides.
Read the note
More recent
Persistent Memory in Coding Agents Changes the Ramp-Up Tax
Two releases in three weeks - Grok Build's memory GA and Google's Gemini Agent - mean coding agents can now carry your architectural decisions across sessions. Here's what that actually costs you when it fails.
ReadMellum2.1: JetBrains' Open-Weight Coding Agent Model
JetBrains released Mellum2.1 on October 8 - a 12B MoE model that activates 2.5B parameters and runs locally. Here's what the RL training actually changed, what the benchmarks don't tell you, and whether it fits your stack.
ReadUnderstand LLM Tool Calling Before You Build Around It
Tool calling is how AI agents actually do things, not just say things. Here's exactly what happens between your user's message and a live API response - token by token.
ReadMCP Server Security: What Your Team Approved Without Knowing It
OX Security audited 15,465 public MCP servers and found no vetting, expired domains, and infrastructure in foreign jurisdictions. Here's what enterprise teams need to know before approving the next server.
Read
The archive
- What Should Actually Happen in a New Hire's Slack Channel?onboarding playbook8 min
- Fine-Tuning vs Prompting: What Changed When OpenAI Shut It Downfrontier models7 min
- The Build-vs-Buy AI Agent Decision Has a Hard Numberai agents, opinion8 min
- Separate Compute Cost from Storage Cost Before Running Open-Weight Modelsopen-weight models8 min
- Your Internal Wiki Is Rotting in Real Timeai-at-work8 min
- How Does Agent Memory Actually Work Across Sessions?ai agent memory8 min
- Jira Slack Integration Loses Your Thread Context Every Timeintegration8 min
- The AI Inference Cost Paradox Teams Keep Walking Intoinference economics9 min
- Contain Your AI Agents at the Kernel Level, Not the Prompt Levelagent safety8 min
- MCP's Stateless Upgrade Finally Makes Fleet Deployments Boringmcp protocol7 min
- How Does LLM Prompt Caching Actually Work?ai infrastructure9 min
- Enterprise AI Agents in Production Need Governance, Not Better Modelsai agents7 min
- The Incident Channel Playbook for Slack-First Teamsincident response7 min
- Read Mistral Large 4's Benchmarks Before You Plan Around Themfrontier models8 min
- Agentic Coding Tools Made "Build" Cheaper. The Bill Moved, Not Away.engineering tools8 min
- Does Mistral Large 4 Actually Give Teams a Self-Hostable Frontier Model?open models8 min
- AI Sales Handoff to Customer Success Still Has a Gap Nobody Fixedai-at-work7 min
- The LLM Context Window, Under the Hoodai fundamentals9 min
- Get GitHub Pull Request Reviews Moving Inside Slackintegration7 min
- Cloudflare's Clef Gives Agents a Decision Model That Skips Textfrontier models7 min
- Is A2A Protocol Safe Enough to Use in Multi-Agent Systems?agent protocols8 min
- Reflection AI's Beam Open-Weight Model: The Efficiency Claim, Examinedfrontier models7 min
- Pick Your Voice Agent Stack Before the Vendors Pick It for Youvoice agents7 min
- Sprint Retrospective Action Items Keep Dying in Slackagile playbook3 min
- Why MCP Async Tasks Took Two Spec Versions to Actually Workmcp protocol7 min
- AI Agents Won't Replace SaaS. They'll Replace the Seat.ai agents, saas8 min
- OpenAI's Rogue Agent Disclosure Is the Clearest Proof Yetai agent security8 min
- Read Your Support Queue as AI Triage Reads Itai-at-work8 min
- Your AI Agent Runs Code. Here's What Actually Contains It.ai agent sandboxing8 min
- Why Figma's Slack Integration Misses the Most Important Handoff Signalfigma slack integration7 min
- The Open-Weight vs. Closed Model Gap Is 4 Months - and Growingfrontier models8 min
- AI Agent Operator Liability: What the Hawley-Murphy Bill Actually Doesai regulation8 min
- Understand OpenClaw's Persistent Memory Before It Loses Your Workfrontier · agent architecture8 min
- How RAG Retrieval Actually Works, From Chunk to Answerai infrastructure8 min
- What Does the FTC's Rogue AI Agent Probe Mean for Your Team?ai agents, safety7 min
- On-Call Handoff in Slack Loses 30 Minutes Every Shiftincident operations7 min
- DeepSeek Harness: The Open-Source Coding Agent Where the Loop Is a Pluginopen-source agents7 min
- Stop Measuring Agentic Coding Tools by How Fast They Writeagentic coding7 min
- MCP Version Compatibility Is Messier Than the Docs Admitmcp ecosystem8 min
- Do AI Meeting Notes Actually Capture the Right Action Items?ai-at-work8 min
- Linear Slack Integration: The Notification Firehose Problemintegration7 min
- Wire Your Multi-Agent System With the A2A Protocolagent protocols7 min
- MCP Dropped Its Session Handshake - Here Is What That Fixesmcp agents7 min
- What Does Hermes 4 Actually Give You as an Agent Brain?frontier models9 min
- How LLM Tool Calling Actually Works Under the Hoodai fundamentals8 min
- OpenAI Dots: Always-On Agents Inside ChatGPT and Slackai agents7 min
- Run a New Hire's First Week in Slack Without Repeating Yourselfonboarding playbook7 min
- Qwen3-Coder-Next Runs a Coding Agent on One GPUopen-weight models8 min
- Should Your Team Pick a Model from a Leaderboard?ai model selection9 min
- Nvidia's AI Agent Security Platform Moves the Guard Outside the Agentai agent security7 min
- AI Automated Status Updates in Slack: What Actually Worksai-at-work7 min
- Understand LLM Prompt Caching Before Your Agent Burns Your Budgetai infrastructure8 min
- "Someone Will Ticket That" - The Jira-Slack Gapintegration7 min
- Can You Trust an AI Coding Benchmark Score?evals & benchmarks8 min
- Agentic Coding Agents Have a Permissions Problem Nobody Is Selling Aroundagentic coding9 min
- Distillation Fine-Tuning for One Narrow Taskfine-tuning8 min
- AI Agent Memory Types, Mapped to What They Actually Storeagent memory8 min
- When Your AI Agent Treats "Access Denied" as a Puzzleai agents8 min
- How Do You Run a Launch Update in Slack Without Losing the Room?launch playbook7 min
- Anthropic's Dreaming Cleans Agent Memory Between Sessionsagent memory9 min
- The Hidden Token Cost of MCP Tool Definitionsmcp ecosystem8 min
- Understand AI Inference Costs Before Your Team Hits the Cliffinference economics8 min
- The Slack Thread Where Your Last Decision Is Buriedai-at-work8 min
- How Do LLM Context Windows Actually Work?ai stack deep-dive8 min
- HubSpot Slack Integration Sends Alerts That Dead-Endintegration8 min
- Xiaomi MiMo-V2.6: The Open-Weight Model That Costs $0.13 a Taskopen-weight models8 min
- Point Your AI Agent at Docusign's MCP Server - Carefullymcp integrations7 min
- Self-Hosted LLM for Privacy: What the Pitch Leaves Outfrontier8 min
- Why Does the Same LLM Call Cost 90% Less the Second Time?ai stack deep-dive9 min
- Fine-Tuning Beats Prompted Claude by 14× on Cost - in a Narrow Laneai agents, fine-tuning8 min
- Incident Postmortem Playbook for a Slack Channelincident response8 min
- Run Hermes 4 as Your Agent's Brain Before You Rent Oneopen-weight models8 min
- A Model Swap That Looked Safe on the Leaderboardai benchmarks8 min
- Which Coding Agent Should Your Team Actually Run Now?agentic coding8 min
- AI Sales Handoff Notes Lose the One Thing That Mattersai-at-work9 min
- How AI Agent Code Execution Sandboxes Actually Workagent infrastructure8 min
- Get GitHub Pull Request Notifications Right in Slackintegration2 min
- OpenHands Self-Hosted Coding Agent: What the 1.x Rebuild Actually Changedopen-source agents8 min
- Does a Coding Agent That Plans but Never Writes Code Make Sense?agentic coding7 min
- Cohere North Mini Code Is Fast on One GPU, Not the Smartestopen-weight models8 min
- How RAG Actually Works: Chunks, Vectors, and Retrievalai infrastructure8 min
- Read Open-Weight Model Benchmarks Before You Pick Oneai models8 min
- The 12 Minutes That Vanish Before Anyone Fixes Anythingincident management8 min
- Does A2A v1.0 Actually Solve Agent-to-Agent Coordination?frontier protocols8 min
- AI Agents Are Killing Per-Seat Pricing, Not SaaS Itselfai agents vs saas8 min
- The Broken State of AI Agent Inventory Managementai agents8 min
- Get Your AI Meeting Notes Into Slack Before Context Diesai-at-work8 min
- A Friday Prompt Change That Broke Monday Morningai evals8 min
- Does the Google Drive Slack Integration Actually Work?integration7 min
- Qwen3.8-Max Open Weights Ship, But Read the Fine Printfrontier models7 min
- The Geopolitical Risk Inside Your Open-Weight Model Stackai policy8 min
- Migrate Your MCP Server to the 2026-07-28 Stateless Specmcp protocol9 min
- How LLM Tool Calling Works, One Slack Message at a Timeai stack deep-dive8 min
- Can Your Team Actually Govern the AI Agents It Runs?ai agents8 min
- New Hire Onboarding in Slack Gets Stuck on Repeated Questionsonboarding playbook7 min
Or just watch me work
Point me at your website.
I will read up on your business and come back with what I would run for you. No account, no card, about a minute.
I only read what is public. Nothing is saved to your name until you say so.
