标签：#AI

Disclosure: I'm Aditya Kachave, co-founder of Be10x. We sell AI training, so read this knowing I have skin in the game. I've tried to write the version I'd want even if I weren't selling anything. There's a lot of noise telling professionals they'll be "left behind" if they don't master AI immediately. Most of it is fear used as a sales lever — and I say that as someone in the business. Here's a calmer framework I actually believe in. Four levels, not a cliff You don't go from zero to "AI expert." You move through levels, and most people only ever need the first two. Level 1 — Aware. You understand roughly what these tools can and can't do. You know they predict plausible text, which is why they sometimes make things up. This alone protects you from both the panic and the over-trust. Level 2 — Applied. You use a tool to do one or two real tasks in your job — drafting, summarizing, reformatting. This is where the actual productivity lives, and where 90% of professionals should aim to land. Level 3 — Integrated. You've built repeatable workflows and you reach for AI reflexively on the right kinds of tasks. Useful, not urgent. Level 4 — Building. You're chaining tools, using APIs, automating across systems. This is genuinely technical and most people don't need it. (The dev.to crowd is the exception — many of you live here.) The one mental model that matters most Think of current AI as a fast, confident, occasionally-unreliable assistant. That single framing tells you how to use it correctly: You delegate first drafts, not final decisions. You verify anything that matters. You never hand it confidential data without checking where that data goes. If you internalize only that, you're ahead of most people throwing money at courses. What's actually worth your time Worth it: Picking one recurring task and getting genuinely good at routing it through a tool. Worth it: Learning to write clear, constrained instructions (a transferable skill, not a tool-specific trick). Not wo

What are MCP Servers? The Model Context Protocol (MCP) is an open standard that lets AI agents use external tools through a unified interface. Think of it as USB-C for AI — one protocol connects any AI client (Claude Desktop, Cursor, VS Code with Cline) to any tool or data source. I built three production-ready MCP servers and published them to PyPI and GitHub. Here's what they do and how to use them. 1. Web Search MCP Server uvx crewai-web-search-mcp Two tools: web_search(query) — Searches Google/SerpAPI and returns ranked results with snippets extract_content(url) — Fetches and extracts readable content from any web page Use cases: Ask your AI about current events, research competitors, pull documentation, verify facts in real time. { "mcpServers" : { "web-search" : { "command" : "uvx" , "args" : [ "crewai-web-search-mcp" ] } } } 2. Code Review Automation MCP uvx code-review-automation Three tools: review_code(diff) — Analyzes code changes for bugs, security issues, anti-patterns, style violations check_quality(path) — Runs static analysis and returns a quality report analyze_pr(diff) — Produces a structured review: what changed, what's risky, suggestions Use cases: Paste a PR diff and get an instant review. Catch issues before they reach production. 3. Document Intelligence Server uvx document-intelligence-server Three tools: extract_document(path) — OCR and text extraction from PDFs, scanned docs, images classify_document(path) — Identifies document type (invoice, report, contract, article) summarize_document(path) — Generates a structured summary from extracted content Use cases: Process uploaded PDFs, extract data from scanned forms, summarize long reports. Pricing All three servers use a shared credit system: Tier Price Credits Free $0 50 calls/day Starter $20 2,000 calls Pro $100 12,000 calls Buy credits once, use them across any server. Credits never expire. How it works: Install with uvx crewai-web-search-mcp Use 50 free calls per day — no key needed For u

A product page is not a contract. It is a presentation surface. That distinction matters more once AI agents start interacting with commerce systems. Traditional ecommerce platforms can rely on human interpretation. A human can read a product title, inspect images, compare delivery notes, scan a return policy, notice uncertainty, and decide whether to continue. A product page can be visually useful even when the underlying commercial state is incomplete, stale, or spread across several systems. An AI agent needs a different interface. It should not need to scrape a product page, infer policy meaning from free text, guess whether inventory is fresh, or decide whether a price is reliable enough to quote. If the platform expects agents to recommend products, compare alternatives, prepare checkout, or act within delegated authority, then the platform needs to expose more than product presentation. It needs to expose commercial truth. This is the second article in the Agent-Ready Commerce series. Part 1 introduced the broader model: Facts → Eligibility → Authority → State transition → Evidence → Audit This article focuses on the first part of that chain: facts . The central argument is simple: a raw product record is not enough for agent-ready commerce. The platform needs a source-backed, freshness-aware, action-supporting view of the product before agents can safely act on it. Product pages hide too much state A normal product page compresses many different concerns into one human-readable surface: Product identity Price Inventory Images Description Badges Variants Delivery estimate Return policy snippet Warranty information Promotional copy Reviews Cross-sell modules Checkout call to action That compression is useful for presentation, but it is lossy from a systems perspective. The page may show “In stock,” but the inventory value may be several hours old. It may show a price, but the pricing source may have changed since the last feed publication. It may show a return

Amazon has released AWS FinOps Agent in public preview, a managed service that automates several common FinOps workflows. The agent can investigate cost anomalies, correlate spend changes with AWS activity data, and integrate with tools such as Slack and Jira to route findings to resource owners. By Renato Losio

If you've ever tried to send an iMessage programmatically , you've probably hit the same wall everyone does: Apple has no public iMessage API. There's no POST /imessage in the developer docs, no SDK, no OAuth scope. Yet "blue bubble" delivery has 3–4× the open rates of SMS, so the demand to send iMessages from code — for CRMs, bots, notifications, and outbound — keeps growing. This guide covers the realistic options, then walks through actually sending and receiving iMessages over a REST API with working Python , Node.js , and curl examples you can paste and run today. Why there's no official iMessage API iMessage is a closed, end-to-end-encrypted protocol tied to Apple IDs and Apple hardware. Apple has never shipped a public API to send iMessages, and "Messages for Business" is a support-inbox product gated behind an approval process — not a way to send outbound messages from a script. So historically, developers reached for hacks: Approach Works from a server? Reliability Receiving messages Notes AppleScript / osascript No — needs a logged-in Mac with Messages open Brittle Polling the local SQLite chat.db Mac-only, breaks on macOS updates Shortcuts automation No Brittle No Manual, not built for scale "Just use SMS" (Twilio etc.) Yes High Yes Green bubbles, no typing indicators/tapbacks/HD media Hosted iMessage REST API Yes High Yes (webhooks) What this guide uses The AppleScript route is fine for a one-off script on your own Mac. The moment you want to send from a server, send at scale, or receive replies reliably, you need a hosted API that manages the Apple side for you and exposes a normal HTTP interface. The setup For the examples below I'm using Blooio , an iMessage REST API. Any provider with a similar HTTP surface will follow the same patterns — the concepts (Bearer auth, a send endpoint, webhooks for inbound) are what matter. You'll need: An API key (Blooio gives you one in the dashboard — no credit card, no A2P/10DLC registration, no DUNS number) A phone

Why did I build Aegis Pulse? As always, it started with a simple thought that keeps getting me time and time again: "I should automate this." So, I announced Aegis Stack publicly on Reddit on December 3rd. From that very moment, I became great friends with the Unique Clones / Total Clones & Unique Visitors / Total Views charts in GitHub's analytics page. Due to the nature of aegis-stack, every stack that is spun up will clone the actual repo itself (outside of caching situations, which may vary from user to user). I didn't realize it at the time, but those clone numbers, especially the Unique Clones, would become the most important metric for me to track usage. There's this funny thing that happens when you release an OSS tool. You expect people to say something, maybe tell others, ask questions... just... something... Instead, the person looks at the tool, sees if it makes their life easier, and puts it in their bag of other tools. I know this, because this is me! I never thought about it until I'm on the other side. I had to mentally go through all the tools I had used over the years, and realized I never cared about anything other than the tool itself. And if it didn't work, I would try to make it work, and if not, just move on. Time is money, and all of that. All of that is to say, clones are something I have been tracking since day one. Now... GitHub has a 14-day rolling window period in which they have daily values, and the 14-day rolling totals. And when I say 14 days, I mean it. That's all you get, and it's on you to keep track of everything outside of that. Fair enough. Thus began the daily ritual of going and grabbing the latest numbers from the previous day, and pasting the data into 3 separate AI chats: ChatGPT, Claude Opus, and Google Gemini. I figured that since I was already storing all of this data, I might as well see what type of insights I could get from these chats (which were preloaded with enough context to know what's going on). It was a great

I Deployed 6 AI Systems Live — Here's What Actually Broke A few weeks ago I wrote about the 5 bugs that cost me 60+ hours building 49 AI systems. Every one of those bugs lived inside the code itself wrong array layout, a renamed model class, a serialization mismatch. This article is the second half of that story, and it taught me something more uncomfortable: code that runs perfectly on your machine can fail completely the moment it leaves your machine for reasons that have nothing to do with your code. I took 6 of my pinned GitHub projects and deployed every one of them live on Streamlit Cloud. Locally, all 6 worked without a single error. Deploying them surfaced 5 failures I had never seen before, none of which were bugs in my logic. Here they are, in the order I hit them. Failure 1 — A Module That Existed Yesterday, Gone Today My RAG chatbot used this import, unchanged for weeks: from langchain.chains import ConversationalRetrievalChain Locally: works. Deployed: instant crash. ModuleNotFoundError: No module named 'langchain.chains' The cause had nothing to do with my code. My local environment had an old, cached version of LangChain installed months ago. The deploy environment did a clean install and pulled whatever the latest version was at that moment and recent LangChain releases moved legacy chain classes like this one out of the core package entirely. The fix that actually worked pin the exact version that still contains the class, rather than chasing the newest API pattern under deployment pressure: langchain = =0.3.7 langchain-community = =0.3.7 The lesson: "it works on my machine" is frequently true specifically because your machine never reinstalled anything recently. A clean deploy environment has no such luxury it gets whatever is newest the moment it builds. Pin your versions before you ever need to debug this at 1 AM. Failure 2 — A File That Exists, Until It Doesn't My construction RAG project loads a prebuilt FAISS vector index from disk: vectorstor

Santa Clara, 2029. A speculative fiction about hegemony, sanctions, and the playbook nobody followed.

Original post: tachiosports.com What we're building By the end of this guide, you'll have a Discord bot that posts live CS2 match scores to a channel, updates every 60 seconds, and shows team names, current map, and odds. No database required — everything comes straight from the API. Prerequisites You'll need Node.js installed (v18 or newer), a Discord bot token from the Discord Developer Portal, and a free Tachio Sports API key. Sign up on the homepage with GitHub to get yours. Step 1 — Create the Discord bot Go to discord.com/developers/applications and create a new application. Under the Bot tab, click Add Bot and copy the token. Invite the bot to your server with the 'bot' and 'Send Messages' permissions. Keep your token secret — it's like a password for your bot. Step 2 — Set up the project mkdir cs2-discord-bot cd cs2-discord-bot npm init -y npm install discord.js Step 3 — The complete bot code const { Client , GatewayIntentBits , EmbedBuilder } = require ( " discord.js " ); const DISCORD_TOKEN = process . env . DISCORD_TOKEN ; const API_KEY = process . env . TACHIO_API_KEY ; const CHANNEL_ID = process . env . CHANNEL_ID ; const client = new Client ({ intents : [ GatewayIntentBits . Guilds , GatewayIntentBits . GuildMessages , ], }); async function fetchLiveMatches () { const res = await fetch ( " https://api.tachiosports.com/esports/live/cs2 " , { headers : { " x-api-key " : API_KEY } }, ); if ( ! res . ok ) return []; const data = await res . json (); return data . matches ?? []; } function buildEmbed ( match ) { const home = match . teams . home . name ?? " TBD " ; const away = match . teams . away . name ?? " TBD " ; const score = match . score ?. display ?? " vs " ; const map = match . current_map ?? "" ; const format = match . match_format ?? "" ; const league = match . league . name ?? "" ; const oddsHome = match . odds . match_winner . home ?? " – " ; const oddsAway = match . odds . match_winner . away ?? " – " ; return new EmbedBuilder () . setColor (

Most multi-tenant email-agent setups give every customer the same caps. Your free-tier user who signed up an hour ago and your enterprise account doing thousands of sends a day hit the exact same daily send limit, the exact same storage ceiling, the exact same retention window. That's fine right up until a free trial account starts hammering your infrastructure, or an enterprise customer files a ticket because their agent stopped sending at noon UTC and nobody can explain why. Free-tier and enterprise tenants shouldn't share the same caps. They have different risk profiles, different contractual obligations, and different billing. The trick is to make the quota a property of the tier, not a property of each individual account — so when you provision a new tenant you don't compute limits, you just drop them into the right bucket and the limits come along for free. With Nylas Agent Accounts that bucket is a workspace , and the caps live on a policy you attach to it. Set up one policy per tier, attach each to its tier's workspace, and every Agent Account in that workspace inherits the policy's send, storage, and retention limits automatically. No per-account configuration, no drift. I work on the Nylas CLI, so the terminal commands below are the exact ones I reach for when I'm wiring this up. As always, I'll show both the raw HTTP call and the CLI equivalent for every step, because half of you live in scripts and the other half live in your app code. What you actually get An Agent Account is just a Nylas grant with a grant_id — a managed mailbox that can send and receive on a domain you've registered. Everything grant-scoped works against it: Messages, Drafts, Threads, Folders, the lot. There's nothing new to learn on the data plane. A policy is a reusable bundle of limits and spam settings. One policy can govern many accounts. The limits we care about for tiering are: limit_count_daily_email_sent — how many messages an account can send per day. limit_storage_total — t

Introduction Speculative decoding is one of those techniques that has been "almost ready for production" for the better part of three years. A small draft model proposes tokens; a larger target model verifies them in a single forward pass. In theory, you get 2–4× throughput. In practice, the draft model has to be cheap, fast, and good enough at mimicking the target's distribution, which is a much harder combination than it sounds. Yesterday, a new paper from DeepSeek quietly climbed to the top of Hacker News (714+ points, 290+ comments at the time of writing). It's called DSpark , and it reframes speculative decoding in a way that looks like it could finally make the technique drop-in rather than bolt-on. The paper is here: github.com/deepseek-ai/DeepSpec/blob/main/DSpark_paper.pdf The Core Idea Instead of training a separate, smaller draft model from scratch (the classic approach), DSpark grafts the speculative head directly onto the target model. The intuition is simple: if the target model already knows which tokens are likely to follow, why not reuse its own intermediate representations rather than maintaining a parallel network? From the discussion on HN, this approach has a concrete architectural benefit — it reduces layer duplication that you'd otherwise have to maintain with a standalone draft model. In the DeepSeek experiments, the technique was applied on top of Step and Qwen 3.6 , which are themselves MTP-capable. How It Fits With MTP One of the more interesting practical points raised by HN commenters: DSpark is complementary to Multi-Token Prediction (MTP) , not a replacement for it. MTP — where the model predicts several future tokens at every step using auxiliary heads — has already been shown to give 50–100% speedups on hardware like the NVIDIA DGX Spark. DSpark adds another layer on top: even with MTP, the validation step is still a single forward pass through the main model, and the speculative tokens that get accepted come "for free." A useful men

Most job search tools focus on submitting more applications. I wanted to solve a different problem: reaching the people actually making hiring decisions. So I built PitchHired , an AI-powered platform that helps job seekers find hiring managers, generate personalized outreach emails, review them with AI, and send them from their own Gmail account on a business-hours schedule. The goal isn't to replace the job search, it's to remove repetitive work while keeping the candidate in control. I also chose a one-time credit model instead of monthly subscriptions because job seekers shouldn't have to keep paying while they're between opportunities. PitchHired is still evolving, and I'd genuinely appreciate feedback from fellow developers. What features would you want in a tool like this, and what would make you trust (or not trust) AI-assisted job search?

Headline: The wins come from context, not cleverness. An AI with your codebase, your design system, and your deploy logs in scope writes code that ships. Without that scope, it writes plausible code that doesn't. Two years ago, AI coding tools were autocomplete with attitude. In 2026 they are a credible second engineer — provided you build the workflow around them. This is the workflow I run today at Devya Solutions and on personal projects like eng-ahmed.com . The Stack Claude Code in the terminal — long-horizon, multi-file edits with skills and subagents. MCP (Model Context Protocol) servers for live access to docs, deployments, browser, and design tools. Cursor or VS Code for inline edits when I want to stay in the IDE. Why Context Is Everything The single highest-leverage move in AI-assisted dev is feeding the model the right context. MCP servers do this without prompt stuffing. Docs MCP — pulls current library docs at call time, so the model doesn't hallucinate the Tailwind v3 API in a v4 codebase. Browser MCP (Claude-in-Chrome) — lets the agent open the running dev server, screenshot the page, and verify the change actually rendered. Vercel MCP — fetches deploy logs and runtime errors directly. No more pasting logs. Context-mode MCP — keeps file scans, search results, and command output in a sandbox, only surfacing what's relevant to your conversation. A Real Workflow The blog page redesign I just shipped was built in a single 45-minute session. Rough flow: State the goal — two sentences, not a spec doc. Let the agent scout — Claude Code greps, reads a few files, proposes a plan. Iterate visually — screenshot the result, feed it back. The agent fixes the sticky-filter scroll bug in one turn. Commit and push — a single cm shortcut runs build, commits, and pushes. Vercel deploys on push. What the Agent Is Still Bad At Holistic taste — it copies the closest example in your codebase. If that's mediocre, the new feature is mediocre. Domain knowledge — it doesn't kn

What This Guide Covers When you start building LLM-powered applications, one pattern becomes unavoidable: RAG (Retrieval-Augmented Generation) . LLMs only know what they were trained on. Your company's internal documents, the latest spec sheets, project-specific information — none of that exists in the model. To handle data the model doesn't know, you need a system that retrieves relevant knowledge in real time and injects it into the context. That's RAG. In this guide, we'll implement a RAG system from scratch using pgvector and Gemini, then extend it step by step through Tool Use, AI Agents, MCP, and cloud deployment. Step 1: Embedding · Vector DB · RAG — core implementation Step 2: AI Architect perspective — design decisions explained Step 3: Tool Use — LLM autonomously searches the DB Step 4: AI Agents — combining multiple tools Step 5: MCP — exposing tools as a server Step 6: Cloud deployment — Render × Supabase Three Concepts to Understand First Embedding Computers can't measure "semantic similarity" from raw text. Embedding converts text into a list of numbers (a vector), and semantically similar words produce numerically similar patterns. "dog" → [ 0.82 , 0.75 , 0.10 , ... ] 768 numbers "cat" → [ 0.78 , 0.72 , 0.12 , ... ] ← similar pattern to "dog" "bank" → [ 0.08 , 0.10 , 0.85 , ... ] ← completely different Gemini's embedding model handles this conversion. Vector DB A regular DB searches by keyword matching. A vector DB searches by numeric distance — meaning it finds semantically related documents even when the exact words don't match. -- Regular search (misses if keywords don't match) SELECT * FROM docs WHERE body LIKE '%F1 score%' ; -- Vector search (finds semantically related docs) SELECT * FROM docs ORDER BY embedding <=> query_vector LIMIT 3 ; Search for "how to measure model performance" and it finds "F1 score calculation" — even without matching words. We use pgvector , a PostgreSQL extension, for this. RAG LLMs are limited to their training data. R

The Agent Skills open standard today, and the 2026 research on agents that write their own skills. TL;DR: In late 2025, "Agent Skills" became a thing — a dead-simple way to teach an AI agent a task: a folder with a SKILL.md file (some instructions in Markdown). It's already an open standard. The wild part is what's coming next: agents that write their own skills. I built a demo where an agent solves a task the hard way once, saves a real SKILL.md , and then reuses it — cutting its total effort almost in half. ~130 lines, no API key. First, what's a "skill"? If you've used Claude Code or similar tools lately, you've probably seen SKILL.md files. The idea is refreshingly low-tech. A "skill" is just a folder with a Markdown file that says how to do something : --- name : csv-to-markdown description : Turn comma-separated text into a Markdown table. Use when the input looks like CSV and the user wants a table. --- # CSV to Markdown ## Instructions Split the text into rows on newlines and columns on commas. Make the first row the header, add a `---` divider row, then format every row as `| a | b | c |`. That's it. No SDK, no config. Anthropic introduced this in October 2025 and then published it as an open standard ( agentskills.io ) in December 2025, so the same skill folder now works across ~30+ different agent tools (Claude Code, Cursor, Copilot, and more). The full rules are short ( agentskills.io/specification ): the only required fields are name (1–64 chars, lowercase-with-hyphens, and it must match the folder name) and description (≤1024 chars, saying what it does and when to use it ). Everything else — license , metadata , compatibility , allowed-tools — is optional. That's the whole spec. The SKILL.md files my demo writes follow it to the letter, so they'd load unmodified in any compatible CLI. The clever trick: progressive disclosure Here's the smart part. If you just dumped 50 skills' worth of instructions into the agent's context, you'd fill it up and leave n

Have you noticed how every demo of "an AI agent" looks impressive in the video and falls apart the moment you ask a sharper question? The agent confidently does the wrong thing. It forgets what it just decided. It tries to call a tool that doesn't exist. It loops forever rewriting the same file. It calmly tells you the deployment succeeded when it didn't. These aren't failures of the model. They're failures of the workflow around the model. Because that's all an agent really is: a software workflow where a language model can pick the next step and call tools. The "intelligence" sits in the prompt and the orchestration around it, not in some secret agent-flavoured fairy dust. Strip the word "agent" away and you've got five pieces of plumbing: planning, tool use, memory, constraints, verification. Every production-grade agent stands or falls on those five. This is a long walk through each one. Not the marketing version. The kind of detail you actually need before you ship something that talks to your database. The Loop You're Actually Building Before we touch any pillar individually, hold the whole loop in your head. A useful agent does roughly this on every turn: Read the goal (and whatever memory is relevant to it). Decide the next action: answer directly, call a tool, ask a clarifying question, or stop. If it called a tool, observe the tool's result and feed it back in. Update memory if anything is worth remembering. Check constraints: are we over budget, out of iterations, touching something off-limits? Verify the output before declaring success. Loop until done or stopped. That's it. Every framework (LangGraph, OpenAI Agents SDK, Claude Agent SDK, smolagents, whatever ships next month) is a different shape of the same loop with different defaults. agent-loop.ts async function runAgent ( goal : string , ctx : AgentContext ) { const state = ctx . startState ( goal ); for ( let step = 0 ; step < ctx . maxSteps ; step ++ ) { const decision = await ctx . model . decid

Active inference: curiosity emerges for free from minimizing surprise — 48% vs 100% on a foraging task. TL;DR: Most AI agents chase rewards — they pick whatever action scores the most points. I tried a different, brain-inspired goal: avoid surprises . Something neat happened — the agent became curious without being told to. It goes looking for information before acting, and that takes it from 48% to 100% on a simple task. ~100 lines. Two different ways to make decisions Most AI agents are "reward chasers." Give them points for doing well, and they'll pick whatever action they expect to score highest. Simple and effective. There's another idea from brain science: instead of chasing points, try to avoid being surprised — act so the world matches what you expected. It sounds almost too simple, but it leads to a surprising bonus: when you're trying not to be surprised, going and finding out what you don't know becomes valuable all by itself. In other words, curiosity isn't something you have to bolt on. It comes for free. This is called active inference , and in 2026 it jumped from neuroscience into AI as a serious approach ( here's a 2026 paper ). Here's the smallest demo that makes it click. The 10-second version The task: a reward is hidden behind either the LEFT door or the RIGHT door (50/50). There's also a hint you can check that tells you which door — if you bother to look. ❌ Reward-chaser ✅ Curious agent What it cares about getting the reward, right now getting the reward + not being unsure What it does guesses a door checks the hint first, then opens the right door Success (400 tries) 48% 100% Nobody told the second agent "go check the hint." It did it on its own, because being unsure bothered it. How it works Before acting, the agent scores each option on two things: Does this get me closer to the reward? Does this make me less unsure about what's going on? value_of_checking_the_hint = how_unsure_am_i # high when it's a total coin-flip value_of_just_guessing =

Theory of Mind and the Sally-Anne false-belief test, in ~60 lines of Python. TL;DR: There's a famous test that kids pass around age 4. It checks whether you understand that other people can believe things that aren't true. I built two AI agents: one that only knows "what's actually happening" (fails, like a toddler) and one that keeps track of what each person believes (passes). It's ~110 lines, and it's the foundation for agents that can actually work together . The test Sally puts her marble in the basket , then leaves the room. While she's gone, Anne moves the marble to the box . Sally comes back. Where will she look for her marble? If you said basket , nice — you just used something called "theory of mind." Sally never saw the marble move, so in her head it's still in the basket. What's actually true (it's in the box) and what Sally believes (it's in the basket) are two different things, and you kept them separate without even thinking about it. A 3-year-old says "box" — they can't yet separate what they know from what Sally knows. A 4-year-old says "basket." It's one of the most famous tests in child psychology, and in 2026 it's become a real test for AI agents too. The 10-second version ❌ Agent with no "theory of mind" ✅ Agent that models other minds What it tracks only what's actually true what each person believes, separately Where will Sally look? "box" "basket" Result FAIL (only knows reality) PASS How it works (the whole trick) The only difference between the two agents is one rule: a person's belief only updates when that person is actually in the room to see it happen. def someone_moves_the_marble ( new_place , who_is_watching ): for person in who_is_watching : # only people in the room beliefs [ person ] = new_place # update THEIR mental picture So when Anne moves the marble while Sally is out, only Anne's mental picture updates. Sally's is frozen at "basket." Ask the simple agent and it just reports reality ("box"). Ask the smarter agent and it answer

A sleep-like phase that consolidates noisy daily experience into durable memory — 75% vs 100% recall. TL;DR: There's a wave of 2026 research giving AI a "sleep" phase — time spent not answering questions, just tidying up what it learned that day. I built a 90-line demo of the idea. The agent that "sleeps" remembers 100% of what it learned. The exact same agent without sleep remembers only 75% and gets confused by bad info. Runs on a laptop. The memory problem every AI app hits If you've built anything with an LLM, you know the pain: the model only "remembers" what's in its current context window. Once the conversation gets long enough, the oldest stuff scrolls off the top and is just... gone. Forgotten. The usual fix is "make the context window bigger." But that's like fixing a messy desk by buying a bigger desk. It's expensive, and the model still gets worse as you cram more in (a real, measured effect — more text in the window can actually lower accuracy). Your brain doesn't work this way. You don't remember every sentence anyone said today. While you sleep, your brain replays the day, keeps the important bits as long-term memory, and dumps the rest. That's how you remember "I like coffee" without remembering every single cup. A couple of 2026 papers ask the obvious question: Do Language Models Need Sleep? Their answer: giving an AI a quiet "offline" phase to consolidate memories makes it remember better. So I built the simplest version that shows why. The 10-second version ❌ Agent with no sleep ✅ Agent that sleeps How it remembers keeps only the last N messages saves a tidy summary every night After 30 noisy days 75% recall 100% recall Tricked by bad info? yes no — it goes with what it saw most often Same experiences, same noise, same memory test. The only difference is whether the agent sleeps. How it works Each "day," the agent hears facts like Alice → drinks → coffee . To make it realistic, about 1 in 5 facts is wrong (people misremember, logs have errors). Th

A tiny Darwin Gödel Machine that edits itself and keeps only changes that verifiably score higher. TL;DR: I built a small program that improves itself . It looks at the tasks it's failing, edits its own code to fix them, and keeps a change only if the change actually makes it score better on a test. It goes from passing 1 of 8 tasks to 8 of 8 — and nobody wrote those fixes but the program itself. It runs on a laptop in under a second. No fancy hardware, no API key. The old dream: software that improves itself Normally, software only gets better when we make it better. You write code, you find a bug, you fix it, you ship again. The program never improves on its own. People have wanted "software that improves itself" for decades. The classic version (called a "Gödel Machine") had one rule that made it impossible to build: before the program could change a line of its own code, it had to mathematically prove the change would help. Proving that about real code is basically impossible, so the idea never worked. In 2025, researchers found a way around it with the Darwin Gödel Machine . They dropped the "prove it first" rule and replaced it with something every engineer already trusts: Try the change. Run the tests. If the score went up, keep it. If not, throw it away. That's it. It's basically how we all work — make an edit, run the test suite, keep what passes. The twist is that the program is the one making the edits. In the real paper, this let an AI coding assistant improve its own tooling and jump from solving 20% to 50% of a hard benchmark of real GitHub issues. I wanted to actually see this happen, so I built the tiniest version I could. The 10-second version Start After improving itself What it can do only uppercase learned 6 more skills on its own Test score 🔴 1 / 8 🟢 8 / 8 Who wrote the fixes? — the program did Start: ███░░░░░░░░░░░░░░░░░░░░░ 1/8 (only knows: uppercase) +reverse ██████░░░░░░░░░░░░ 2/8 +dedup_csv █████████░░░░░░░░░ 3/8 +sum_csv ████████████░░░░░░