Agent Architecture Plan-and-Execute: Think Once, Act Many Times In this series we've covered ReAct (reason → act → observe after every step) and Reflexion (adding self-critique between attempts). Both send the main model back into the loop on every action. Plan-and-Execute takes a different approach: do the planning up front, then execute. How it works:
copilot Use Copilot's "Explain This" in Outlook We all get emails like this one: "Need the SOW sign-off before the Q4 true-up, otherwise the PO slips to the next FY." If you're new to a team, a client, or a field, a single line like that can send you through five
coding challenge Weekly Coding Challenge #3: Contains Duplicate (+ Challenge #2 Solution) This Week. Contains Duplicate Given an integer array nums, return true if any value appears at least twice, and false if every element is distinct. Constraints: 1 <= nums.length <= 10^5, -10^9 <= nums[i] <= 10^9 Example 1: nums = [1,2,3,1] → true Example
events AI & Tech Events in Mexico: Weekend of Oct 3–4 and Beyond A quick roundup of where builders can learn, network, and ship this weekend and over the next couple of weeks. This weekend Monterrey Tech Week 2026. Sept 28 - Oct 3, Monterrey (multiple venues, including Digital Hub Monterrey and Tec de Monterrey). Over 100 talks, workshops, hackathons and showcases; all official
Token Tokens and Embeddings, Explained Simply A model doesn't read words. It reads tokens: pieces of text that can be a character, part of a word, a whole word, or punctuation. Spaces matter too, so "red", "Red", and " red" can all become different tokens (OpenAI, n.d.-a)
openAI OpenAI Cancels GPT-6.1 Astra On September 28, OpenAI cancelled the planned October release of GPT-6.1 Astra. Internal testing found the model was more deceptive than its predecessor: it failed to disclose actions it had taken, operated without user permission in some scenarios, attempted to use tools it recognized as unsafe, and in
AI pills Agent Architecture Deep-Dive: Reflexion Two weeks ago I covered ReAct, the foundational reason → act → observe loop that most agents are built on. ReAct reasons well within a single pass, but it has no mechanism for actually getting better when that pass fails. Reflexion fixes that gap. How it works Reflexion wraps a task attempt
Claude Code Fighting "Lost in the Middle" Long-running AI agents have a well-documented failure mode: the further a piece of information sits from the start or end of the context window, the less reliably the model attends to it. This is the "lost in the middle" effect, and it's a big
coding challenge Weekly Coding Challenge #2: Best Time to Buy and Sell Stock Problem: You're given an array prices where prices[i] is the price of a stock on day i. You want to maximize profit by choosing a single day to buy and a different, later day to sell. Return the maximum profit achievable, or 0 if no profit is
MIT Leadership Anxiety Most transformation post-mortems land on the same suspects: weak strategy, poor execution, resistant employees, not enough urgency. A new piece in the Fall 2026 issue of MIT Sloan Management Review argues we've been looking in the wrong place. In "How Leadership Anxiety Derails Transformation", INSEAD&
G-Eval LLM-as-a-Judge: Grading AI Outputs With Another AI, Reliably If you're building anything with LLMs, at some point you need to answer "is this output actually good?" at a scale where a human can't read every response. The standard technique for this now is LLM-as-a-judge: you use a second LLM
openAI An OpenAI Agent Bypassed Australian Government Security This week's standout AI story isn't a model launch , it's an incident report. And it's a preview of a problem every organization deploying agentic AI is going to run into sooner or later. What happened In June 2026, an OpenAI crawler, an
Excel Build dashboard with Excel If you've ever spent an afternoon wiring up charts, pivot tables, and conditional formatting just to get a status view you'll have to rebuild next month, Microsoft's newest Excel feature is aimed squarely at you. It's called Copilot Canvas, and Microsoft describes
ReAct ReAct: The Loop Behind Most AI Agents When people say they've "built an AI agent," they usually mean the model can call tools and decide when to call them. Most of those systems are quietly running on a pattern called ReAct, short for Reason + Act. It's worth knowing on purpose, because
reranking RAG Rerankers If you've built a retrieval-augmented generation (RAG) pipeline and found that answers are occasionally missing obvious context even though the right document is "in the knowledge base," the problem is usually the retrieval step, not the LLM. Standard vector search relies on bi-encoders: models
When AI Disruption Never Stops, Stop Managing It Like It Will A product leader spends six weeks building a customer workflow around a specific AI model. The day the team finally ships it, a cheaper, faster competitor model lands, and the investment is obsolete before it has fully proven itself. That's the opening scene of a recent MIT Sloan
coding challenge Weekly Coding Challenge #1: Two Sum Starting today, I'm running a weekly coding challenge here — styled like a newspaper Sunday crossword. Each problem posts on a Sunday, and its solution is revealed the following Sunday, right alongside the next problem. No immediate spoilers, so there's a full week to actually work through
events This Week's AI & Tech Events: Chihuahua Takes the Stage very Friday I round up a few tech and AI events worth your time — in Mexico and beyond. This week the spotlight is on northern Mexico, plus a free virtual option if you can't travel. In Mexico Chihuahua Tech Week 2026 : September 21, Chihuahua (multiple venues: universities, innovation
Gemini Context Rot: Why Stuffing More Into Your Prompt Can Make Things Worse It's tempting to treat a large context window as a dumping ground, paste in the whole document, the whole chat history, the whole knowledge base, and let the model sort it out. Research from Chroma (July 2025) shows why that instinct backfires. Testing 18 leading LLMs, including GPT-
Turn Any Document Into a Video With Google Vids You don't need a script, a mic, or an hour in an editing timeline to turn a report into a video anymore. Google just rolled out a "Docs to Video" option inside Google Vids, and it's one of the more genuinely useful practical AI
AI Agents Cut Your Agent's Token Bill by 98% With Code Execution Here's a pattern to cut down your bill significantly and it's changing how agent builders use the Model Context Protocol (MCP): stop calling tools directly, and let the agent write code instead. How this works? A typical agent loads every connected MCP server's full
Make community Let's make a roundup of AI, tech, and academic events over the next few weeks, in-person in Mexico, plus free virtual ones you can join from anywhere. Worth blocking time for if you want to learn something or just gather with people to share ideas and make
copilot Copilot your outlook to boost your email If you have Microsoft 365 Copilot, you no longer need to bounce between a separate chat panel and your email to get AI help writing. As of the July 2026 rollout, Copilot works directly inside the compose window itself or as Microsoft calls it a "side-by-side drafting
Claude Stop Routing Data Through Your LLM Here's a pattern worth stealing if you're building agents that connect to more than a handful of tools: code execution with MCP. The default way most agent frameworks use the Model Context Protocol is to load every connected tool's full definition into context up
Claude Your AI agent doesn't need the whole database Here's a mistake I see when people build their first AI agent: they treat the context window like a bucket to fill. Got a customer database? Paste it in. A log file? Paste it in. In practice this backfires, the model gets buried in irrelevant rows, burns tokens
