A technical breakdown of what actually happens inside an AI agent — planning loops, tool calls, memory, and verification — aimed at builders who are tired of demo-ready agents that collapse in production. Solid reference material for anyone building agent infrastructure rather than agent demos.
Explore the radar’s memory
Every pick we’ve ever shipped, plotted as a living constellation. Picks cluster by topic; threads between clusters reveal cross-topic connections. Zoom in to expand a cluster into its picks — drag, filter, click through to the brief.
// Browse
All picks
Archive results
-
Inside An AI Agent: Planning, Tool Use, Memory, Constraints, And Verification
-
VP of Nothing: The CEO's Nephew Took Over My AI Platform. The Client Walked Within a Month.
A firsthand account of watching an AI platform project collapse after organizational politics handed the keys to the wrong person. Useful as a cautionary read for anyone building internal AI tooling inside a client's org rather than their own.
-
Thank you DEV community: the Thinking Engineer Toolkit is live
A curated collection of mental models and frameworks for engineers who want to think more clearly about technical decisions. More of a reference resource than a tool. Useful if you are a solo founder who finds yourself making the same architectural mistakes twice.
-
Stratagems #1: Mark Johnson Walked Into an AI Audit. The Benchmark Had Everything Figured Out — Except the Truth.
A narrative-format essay arguing that AI benchmarks breed complacency by making auditors feel the hard work is already done. More editorial than technical, but the core point about over-reliance on benchmark scores in compliance contexts is worth thirty seconds of a founder's attention.
-
KV Cache in LLMs: The Optimization That Makes Modern AI Models Feel Fast
A plain-language explainer on KV caching in large language models — what it is, why inference feels faster with it, and how it works mechanically. Useful background reading for engineers integrating LLM APIs who want to understand the latency behavior they are observing.
-
Our Competitor Had an AI That Covered 97.2%. We Had a Spreadsheet and a Fake Quote. Guess Who Won.
A first-person story about winning an enterprise RFP against a heavily AI-tooled competitor using old-fashioned sales craft. Useful counter-programming if your team is over-investing in capability theater instead of customer understanding.
-
When Software Started Writing Software: A Developer’s History of AI
A developer-written retrospective on how AI coding tools changed the job over the last three years. More essay than tool, but worth skimming if you want a grounded narrative rather than hype to share with a skeptical teammate or client.
-
Turing's Mirror - A Game About the Question We Still Haven't Answered
A game jam entry built around the Turing Test concept, asking players to distinguish AI from human responses. More of a creative experiment than a tool, but relevant if you are thinking about how to design AI evaluation experiences.
-
(new) Bifrost Edge: MCP Visibility and Control for Enterprise Teams and Beyond 🔥
Adds an observability and policy-enforcement layer on top of MCP servers — who is calling what tool, with what arguments, and whether that call should be allowed. Aimed at teams running MCP in production who have started worrying about what their agents are actually doing.
-
A Company AI Flagged My Article As "Low Quality." I Ran the Numbers. Then I Ran Again.
A first-person account of getting flagged by an AI content moderation system, then auditing 347 flagged posts to understand what the model was actually measuring. Worth skimming if you are building or evaluating any AI quality-scoring layer in a content pipeline.
-
AI Agent Memory: Conversation vs Context
A conceptual breakdown of the two memory modes in AI agents — semantic conversation history versus exact context references — with a practical example using Strands and AgentCore. Aimed at engineers building agents who keep running into retrieval or coherence bugs.
-
How I use premortems with Claude and Codex
A practical walkthrough of running structured premortems inside Claude and Codex before shipping features. Useful if your current AI-assisted code review feels like a rubber stamp rather than a real adversarial check.
-
I Spent 3 Months Training An AI. My VP "Reallocated" It. Then I Got Two Calls At 1 AM.
A firsthand account of training an alert classification model that cut false-positive alarms from 60 percent to 7 percent, then watching it get defunded — until a production incident made the case for it overnight. Useful read on internal AI adoption dynamics more than on any specific tool.
-
I Got Flagged by Sloan. Sloan Is a Guy I Know.
A first-person account of getting a piece flagged as AI-generated after explicitly writing about why AI detectors are unreliable. Worth reading if a founder is thinking about AI content policies or building anything in the detection or trust-and-safety space.
-
My AI Agent Hit a Login Wall: BrowserAct Let It Ask for Help and Resume
A writeup on BrowserAct, a library that lets a browser automation agent pause on auth walls, hand control back to a human, then resume. Solves a real pain point in agent workflows where unattended sessions break on login prompts.
-
Building a Chrome Extension to Make AI Use More Intentional
A walkthrough of building a Chrome extension that adds friction to AI tool usage, prompting the developer to pause before reaching for an LLM. More a reflective dev essay than a finished product, but worth five minutes if you are thinking about how your team uses AI day-to-day.
-
My daughter asked if developers used to write code by hand, but it was the follow-up question that surprised me.
A first-person essay from a developer about their child learning to code through AI tools, framed as a reflection on what programming skills mean in an era when eleven-year-olds can vibe-code. Interesting as a think-piece, not as a tool.
-
The 'Prompt' Is Not a Skill — And We Need to Stop Pretending
An opinion piece arguing that prompt writing is not engineering and that treating it as a specialized skill overstates what is essentially just describing a task. The argument is not new but lands some useful points about where actual AI engineering value sits.
-
AI Usage Statistics 2026: The Structural Shift Behind Adoption, Work, and Hiring
A statistics roundup on AI adoption, workforce impact, and hiring patterns as of 2026. Useful as a reference document when you need numbers for a deck or a hiring argument, but it reads more like a market-research summary than an analytical piece.
-
Your Logs Have the Answer. You Just Can't Find It Fast Enough.
A post walking through how one team used structured querying and filtering to cut incident root-cause time during a checkout outage. More case-study than tool, but it surfaces practical patterns for teams still grep-ing raw log streams during incidents.
-
From Prototype to Production: Finishing Moonsu Link, a Chat-Native Agricultural Marketplace for Cameroon
A write-up covering the build-out of a chat-native agricultural marketplace for Cameroon, using conversational interfaces as the primary UX layer. More of a case study than a deployable tool, but the chat-as-marketplace framing for low-bandwidth markets is worth a read.
-
60 Billion into AI: The Final Mile of Xiaomi AI Ambition
A writeup breaking down Xiaomi's 60 billion yuan AI investment plan, with a focus on what the hardware-to-AI stack play means for device-native AI. More useful as competitive context than as a tool, but worth skimming if you are building anything in the on-device AI space.
-
Mass layoffs caused by AI
A personal essay tracking the shift from AI-layoff speculation in 2024 to documented cases in 2025 and 2026. No tool here, but a useful grounding read for founders who want concrete examples of where AI displacement is actually happening versus where it is still hype.
-
AI Native DevCon Day 2: From Agent Demos to Operating Models
A recap of a conference day focused on moving AI agents from demo-stage to operational discipline — covering evals, reliability, and team structure. Useful if you are building toward production agents and want a condensed summary of what practitioners are currently thinking.
-
Every tool seems to have a coding agent horned in these days..... I don't think that makes sense.
An opinion post questioning whether bolting a coding agent onto every developer tool is actually useful, or just competitive feature-stuffing. Short read, worth skimming if the proliferation of half-baked coding agents in your toolchain is already annoying you.