标签：#ai

Most multi-agent frameworks for software development organize agents around roles : a product manager agent, a developer agent, a tester agent. ChatDev and MetaGPT pioneered this approach, and it works well for monolithic tasks. But I ran into a wall when I tried to apply it to a real system with multiple independently-deployed services. The Problem with Role-Based Coordination Imagine you have a backend search service and a frontend management console. The backend team implements a new API endpoint. The frontend needs to adapt. In a role-based framework, there's no natural mechanism for this. Both agents are "developers" in the same simulated organization. There's no concept of service boundaries, no versioned contracts, no way to say "the backend changed, and the frontend needs to know exactly what changed." The coordination problem in multi-service development isn't "which role should handle this task" — it's "which service needs to know about this change, and what exactly changed." That reframing led me to build something different. AgentNexus: Coordinating Agents at the Service Granularity AgentNexus is a document exchange center that treats each service as a first-class citizen. Instead of roles, it uses service boundaries as the coordination primitive. Here's how it works: Each service registers as a sub-project with its own document namespace Services publish versioned Markdown documents: requirements, design specs, API docs, config Services subscribe to documents from other services they depend on When a subscribed document changes, the subscriber receives a diff-aware notification containing both the structured diff and the full latest content The whole thing is exposed as an MCP (Model Context Protocol) server running in streamable-HTTP mode, so multiple agents can connect simultaneously from different machines. The Diff-Aware Update Protocol This is the part I'm most proud of. When an agent calls get_my_updates_with_context , it gets back: { "update_id"

Hey everyone 👋 I'm exploring ideas for an AI-powered developer tool, but before building anything, I want to understand the real problems developers face every day. There are already plenty of tools that generate code. What I'm interested in is everything around coding: Debugging Code reviews Technical debt Documentation Dependency upgrades Testing Deployment Architecture decisions Learning large codebases I'd love to hear from you: A few questions: What's the most frustrating part of your workflow? What task takes more time than it should? What's something you wish AI could do for you today? Have current AI tools (ChatGPT, Claude, Cursor, Copilot, Gemini, etc.) failed you in any important way? If you could eliminate one developer headache forever, what would it be? I've also created a short 2-minute survey: 🔗 https://docs.google.com/forms/d/e/1FAIpQLSf1M5d2y-0RXEIhrbDBtS5gC900YuzWl43cJCxGUrU38MyeDQ/viewform?usp=publish-editor I'll happily share the survey results and key findings with the community once I collect enough responses. Thanks in advance for any feedback!

NVIDIA announced the new Isaac Gr00t reference design humanoid robot platform that combines humanoid robot, five-fingered hands and NVIDIA Jetson Thor compute.

The use of AI Agents creates a distinctive smell... One can tell the GH Repo owner was high on...

How to make LLMs deterministic, in plain English. The version I share with founders and product teams before they make decisions worth real money. You use AI tools every day. But can you explain what happens when you hit send? Most people cannot. And that gap is costing them. Bad prompts. Broken products. Decisions made on the wrong assumptions. The Hard Truth Every LLM explainer out there is written for researchers or so basic it tells you nothing useful. Neither helps you build better products or work with AI more effectively. This is the version I share with senior leaders, founders, and product teams before they make decisions worth real money. 1. It Is Not a Search Engine. It Is Not a Database. It Is a Prediction Machine. When you type a prompt and hit send, the LLM is not finding an answer from somewhere. It is predicting the most likely words to follow your input. Based on patterns it learned from billions of documents. That is the whole process. Wrong: "The AI knows the answer." Right: "The AI predicts the most likely answer based on what it has seen." This changes everything about how you use it. When an AI gives you a wrong answer confidently, it is not broken. It is doing exactly what it was built to do. Predict. Not verify. 2. The Autocomplete Comparison (And Why It Only Gets You Halfway) You have probably heard the phrase "autocomplete on steroids." It is not wrong. But it misses something important. Your phone autocomplete learned from your messages. An LLM learned from most of the written internet. Books. Research papers. Code. Billions of examples. At that scale, the patterns start to look a lot like real thinking. Not because the model understands in the way you do. Because it has seen so much that it can predict what a good answer looks like. When I was building AstroNayak I fed Vedic astrology principles into the system prompt. The LLM produced interpretations that genuinely surprised me. It did not know Vedic astrology. It had seen enough of it t

The row looked perfect. rating: 7 . Valid JSON, right type, no nulls, no missing keys. My schema check waved it through. The page had returned HTTP 200. The selectors hadn't moved. Everything green. A rating of 7 on a 5-star site is impossible. The model invented it, formatted it correctly, and handed it to me with total confidence. That's the failure I want to talk about. Not the scraper that breaks loudly. The one that hands you a clean-looking row that is quietly, plausibly false — and sails past every check you have, because your checks are all looking at the shape of the data, and the lie is in the value . TL;DR HTTP 200, intact selectors, and valid JSON tell you the form is fine. They say nothing about whether the value is true. When an LLM extracts from messy free-text, structured-output mode guarantees you get valid JSON. It does not guarantee the content is real. The model fills uncertain fields rather than leaving them empty — because the schema demands a complete row. A ~60-line value-level sanity gate (ranges, dates, cross-field, reference, language) catches the obvious lies before they hit your database. Real code and real output below. The honest catch: this gate catches rule violations , not plausible lies inside the allowed range . A rating: 4 where the truth is 2 slides right through. I'll be specific about where the gate stops. Two different ways a scraper lies to you I wrote about source drift last week — the case where the page changes underneath you and a 30-line schema check catches the structure shifting. That's an input problem. The source mutated; your agreement with the page broke; you detect it by watching the shape. This is the other end of the pipe. The source is fine. The page is intact, the selectors are correct, the structure is exactly what you expected. The thing that lied to you is the model , on the extraction step, when you asked it to pull structured fields out of a paragraph of human prose. Those two failures feel similar and t

Based on real system architecture decisions. About a $660K AI platform, three AI agents that kept the dashboard green, and a P0 incident that cost $3.15M over one weekend. Act 1 · The All-Hands Meeting Wang Lei, VP of Product, stood in front of the big screen, a smile on his face. Behind him, a dashboard rolled data from the "Axon AI Client Engineering Platform — Q1 Performance Report." Numbers cascaded across the wall: Metric Axon Platform Human Team (Last Q1) Improvement Avg daily tickets processed 847 312 +171% Avg first response time 12s 4h 17m ↓ 99.92% Customer satisfaction 4.8/5 4.1/5 +17% Monthly operating cost $52K $133K −61% Twelve department heads sat in the room. Dead silence. Wang Lei planted both hands on the table and scanned the room. His eyes landed on me. "Alex. Your team processed 312 tickets last Q1. Axon processed more than that in a single day last month." He smiled. Not a friendly smile. A sentencing smile. "And Axon costs less than a third of your team's operating expense." "We invested $660K in the whole platform. At current operating costs, it pays for itself in eighteen months." "After management review — the Client Engineering technical liaison function is being fully transitioned to the Axon platform." He clicked to the next slide. "Employees in replaced roles will complete exit interviews within the week." Someone inhaled sharply. I didn't. I opened my notebook to page 37. "Wang, what dimensions are these numbers from?" "What do you mean, 'what dimensions'?" His smile tightened. "Of those 847 daily tickets — how many are auto-tagging and routing, and how many are actual technical resolutions?" The room went quiet for about five seconds. Wang Lei looked at me. "Axon's ticket closure rate is ninety-three percent." "What's the reopen rate?" He paused. "What?" "After Axon replies — how many customers reopen the same ticket within twenty-four hours?" "We're still collecting that —" "Let me save you the trouble." I turned my notebook toward th

Last year I needed to ask for a raise. I knew my number, I'd read the guides, I had bullet points in my notes app. Then my manager said "let's chat about your goals for next quarter" and I said "sounds great, looking forward to it" and hung up. Never brought up money. Same thing kept happening elsewhere. Coworker taking credit for my work, I said nothing. Relationship that should've ended months earlier, I kept postponing. I always knew what to say. I just couldn't say it with someone actually looking at me. So I started building a thing to practice on. That thing became cosskill . What it actually is You pick a persona, tell it the situation in a sentence, and start talking. The persona doesn't help you. It holds position and pushes back. You practice not folding. Think of it as a flight simulator for hard conversations. You rehearse until your opener comes out steady, then go do the real thing. 20 personas across five categories: Operators (Musk, Jobs): first-principles thinking, harsh product feedback Strategists (Trump, Buffett): treat everything as a deal or a bet Relationship (Ex, Coworker): breakups, workplace friction, family money Philosophy (Socrates, Aurelius, Confucius, Sun Tzu, four more): each tradition frames problems differently Psychology (Rogers, Rosenberg, Ellis, Frankl, Kahneman, Jung): therapeutic frameworks on real situations These aren't celebrity impressions. The Buffett persona won't hype your startup idea. It'll ask "what's the downside?" and keep asking until you have something concrete. Tech stack Next.js 16 on Cloudflare Workers. DeepSeek for inference. Cloudflare D1 (SQLite at edge) for the bits that need to persist. No user accounts, chat history lives in localStorage. Monthly cost stays low enough that the free tier (10 messages/day) doesn't worry me. Why I made these choices DeepSeek instead of GPT-4/Claude. Each conversation is 10-30 messages. At GPT-4 pricing a free product bleeds money. DeepSeek gives maybe 90% of the quality for

There is a specific failure mode in AI-assisted QA work that most tooling discussions skip entirely, and it shows up earliest when you are working solo on a real engagement. Every new chat session is stateless. You paste the ticket, describe the feature, explain your severity logic, set up the context, and by the time the AI is actually useful, you have rebuilt your methodology from scratch for the third time that week. That is not a workflow problem you fix with better prompts. It is an architecture problem, and the fix is a skill file. QAJourney has a full breakdown of this system at qajourney.net/ai-qa-workflow-for-real-projects, including the actual skill files as free downloads. The short version: a skill file is a context document you load as a system prompt. It carries your test surface tiers, your three-path testing framework, your bug report format, your severity and priority logic, your Playwright conventions, and an explicit definition of what the AI does and does not get to call. Load it once per session. The AI operates inside your methodology from the first message instead of a blank slate. The local LLM layer solves a different problem. On a freelance or retainer engagement, tickets contain real product logic and real client data. Sending that to a cloud API on every session is a data exposure question whether or not it rises to a compliance issue. Running Ollama locally with the same skill file as system context keeps the engagement data on the machine. For the output quality required on QA tasks, current 7B to 14B models are sufficient. The cost at zero marginal per token makes it infrastructure rather than a service you pay by the session. The three-role setup in the workflow: engineer as judgment layer, cloud AI loaded with the skill file for complex reasoning and active session output, local LLM for lightweight tasks and client data work. The skill file is the constant across all three. The part that took time to internalize: AI dev teams already

RecursiveMAS (arXiv 2604.25917) showed that agents sharing internal reasoning state outperform agents that share only final outputs. The average accuracy gain across benchmarks was 8.3 points. The mechanism: each agent passes not just its answer but the latent embeddings from its own reasoning process, and the next agent conditions on both. The paper is a good result. The catch is access. RecursiveMAS requires open-weight models with hidden states exposed at inference time. That rules out Claude, GPT-4o, and Gemini. I built a Claude-native version using the Anthropic extended thinking API. The core idea transfers: instead of passing latent vectors, pass the full thinking text. The paper calls it internal state sharing; the Claude version calls it thinking-block relay. The architecture problem Claude's extended thinking blocks carry an encrypted signature tied to the originating conversation. You cannot pass a signed thinking block into a different agent's messages array. The API rejects it. The workaround: extract the text from the thinking block and inject it as a regular user message. # Extract thinking text from Agent 1 thinking_text = next ( ( b . thinking for b in response . content if b . type == " thinking " ), "" ) # Inject into Agent 2 as regular context, not as a thinking block context = f " Prior agent reasoning: \n { thinking_text } " The signature does not transfer. The reasoning does. relay-structured: what I built first The first architecture was a Planner > Critic > Solver loop where each agent emits a compact mental model JSON instead of raw thinking text. Raw thinking at a 1024-token budget is often compressed and fragmented. The hypothesis was that 150 tokens of structured signal carries more information per token than 1024 tokens of compressed prose. The schema each agent emits: { "interpretation" : "how the agent read the problem" , "key_steps" : [ "step 1" , "step 2" ], "rejected_approaches" : [ "approach tried and discarded" ], "confidence" :

I run Sola, a travel app for people who move differently from the traveller the industry was built for. While building it, I kept hitting the same wall. The data I wanted to query did not exist. Not because nobody collected it, but because the schema underneath the whole industry never had a field for it. So on 27 May 2026 I sat down and audited Booking.com. The homepage form, the currency selector, a Bangkok search results page. I wrote down what it accepts and what it refuses. Then I looked at the new AI travel agents shipping on top of it. Here is what I found, and why it matters to anyone building in this space right now. The form is the spec Booking.com's homepage search bar accepts exactly four inputs: A destination, as a single text field A check-in date and check-out date, as one range An occupancy counter, defaulting to "2 adults · 0 children · 1 room" A search button That is the spec. An online travel agency (OTA) is a CRUD app over this spec, and Expedia, Agoda, and Hotels.com run the same four fields. Airbnb lets you skip the dates. The destination stays a single field everywhere. Think about what a spec encodes. The default occupancy is a couple. Not a solo traveller, not a parent with one child, not three generations, not seven people eating from one host's kitchen. The form cannot accept a circuit ("Bangkok, then Hanoi, then Jakarta" forces three separate searches). It cannot accept an open date ("October, not sure which week"). It has no field for the part of a trip where you sleep at family but spend money in restaurants. When you fill that form, you have not searched. You have submitted to a schema. Most of the world's travellers fail the schema before they fail the search. The data receipts I am a builder, so I went for counts, not adjectives. Everything below rendered on the platform on 27 May 2026. Currencies: 52 offered, about 180 in circulation. Eight currencies sit featured at the top of the dropdown. On the day I ran it the order was EUR, US

System-prompt caching alone cut repeat-call costs by half Tool definitions cache separately, perfect for agent loops Conversation history caching pays off after turn three 1-hour TTL beats the default 5 minutes for batch jobs My Anthropic API bill dropped 70 percent last month and I did not change a single model. I changed where the cache breakpoints went. Here are the five patterns I now use on every Claude integration I ship. Pattern 1: Cache The System Prompt First The system prompt is the cheapest win and most people skip it. My agents run with a 4,000 token system prompt that explains the role, the output format, the safety rules, and a few examples. That prompt never changes inside a session. Before caching, I paid full input price for those 4,000 tokens on every single call. With an agent that loops 30 times to finish a task, that is 120,000 tokens of pure repetition. The fix is one parameter. I add a cache_control block with type: "ephemeral" to the last content item in the system prompt array. The first call writes the cache and costs slightly more (cache writes carry a small premium). Every call after that reads the cache at roughly one tenth the input price. Here is the rule I follow: the cached block has to be at least 1,024 tokens for Claude Sonnet, or it gets ignored silently. My 4,000 token prompt clears that easily. If your system prompt is short, this pattern does nothing, so do not bother adding the breakpoint to a 200 token instruction. The order matters more than people expect. The cache works as a prefix. Everything before the breakpoint gets stored. Everything after it is read fresh. So I put the stable stuff (role, rules, examples) up top and the volatile stuff (user query, current date) down below the breakpoint. Reorder this wrong and your cache hit rate collapses because the prefix changes on every call. One real number from my logs: a document-classification job that runs 2,000 times a day. The system prompt is 3,800 tokens. Caching it sav

Claude Code — My daily dev tool Claude Code by Anthropic is the one I use the most for development, by far. What sets it apart from the others: it integrates directly into the terminal and editor, it can read and modify files, navigate an entire codebase, and understand the global context of the project. Not just responding to a copy-pasted snippet in a chat window. In practice, when I have an idea, I ask it to structure the project and challenge my choices. And to be clear: I challenge it too. 😄 I sometimes disagree with its suggestions, and that's often where the conversation becomes interesting. It's a tool, not an oracle. Perplexity — My reference for research Perplexity is my main tool when I need a reliable and verifiable answer. It's a response engine that systematically cites its sources — you ask a question, it answers with excerpts from real web pages and direct links. No more hallucinations without references. However, I use it almost exclusively on desktop. On smartphone, it's flooded with messages pushing the paid version. Understandable from their side, but frankly annoying when you just want to do a quick search. 🙄 Gemini — For those in the Google ecosystem Gemini is Google's AI, and its main advantage is integration with Gmail, Docs, Drive, Sheets, and Google Search. I have a Google Pixel, and on that side, it does integrate very well with its own ecosystem. It's practical for analyzing documents or getting a quick summary without leaving the interface. That said, in terms of responses, it sometimes falters. 😬 Not systematically, but regularly enough that I stay on guard. And if privacy is a priority for you, it's worth thinking twice before entrusting it with your documents — I talk about this in my article on securing yourself on the Internet . ChatGPT — The natural entry point ChatGPT by OpenAI is the most known and most versatile AI. Writing, code, analysis, translation, summary, creativity... it does a bit of everything, often very well. The fre

I do a fair amount of competitor backlink research, and the workflow always annoyed me: open a dashboard, run a query, export a CSV, eyeball it, copy domains into a doc, switch to email. Lots of tab-hopping for what is fundamentally a data-filtering problem an agent should handle. So I wrapped the backlink API I'd been using into an MCP server. Now I stay in Claude Code (or Cursor, Cline, Zed, Windsurf) and just describe the goal. This is the build: the architecture, the four tools, and the one design decision I'm still not sure about. The data source The server runs on the Common Crawl hyperlink webgraph — about 4.4 billion edges across 120 million domains, published quarterly as Parquet. That matters for an MCP tool specifically: the data is open, so there's no scraped-proprietary-index liability in handing it to an agent, and the same query is reproducible by anyone. The HTTP API in front of it ( CrawlGraph ) does the heavy DuckDB work; the MCP server is a thin TypeScript stdio client over it. Keeping the server thin was deliberate — all the query cost, caching, and quota logic lives server-side, so the MCP package stays a ~300-line wrapper that's easy to audit before you hand it your API key. The four tools backlinks → referring domains for a target, with authority scores gap_analysis → domains linking to your competitors but not to you gap_outreach_targets → the composite play (below) releases → list the Common Crawl snapshots backlinks and gap_analysis map 1:1 to API endpoints. gap_analysis is the interesting primitive: submit your domain plus 2-5 competitors, and it returns every domain that links to at least one competitor but not to you, each tagged with a found_on array listing which competitors it links to. The composite tool, and the decision I'm unsure about Most API-wrapper MCP servers are pure 1:1 mappings. I added one opinionated composite tool, gap_outreach_targets , because the raw gap output isn't the thing you actually want — it's the raw materia

LangGraph Production, RAG Memory Challenges, and AI Agent Patterns Today's Highlights Today's highlights dive into practical LangGraph pipeline construction for agentic AI workflows, reveal critical insights from real-world RAG retrieval failures, and unveil 29 open-source design patterns for building robust AI agents. Building Your First LangGraph Pipeline: A Decision-Maker's Guide (Dev.to Top) Source: https://dev.to/labyrinthanalytics/building-your-first-langgraph-pipeline-a-decision-makers-guide-4e25 This article serves as a comprehensive guide for developers looking to implement their first LangGraph pipeline for agentic AI workflows. LangGraph is highlighted as a leading framework for building complex, stateful multi-actor applications, particularly valued for its production readiness and active maintenance. The guide aims to demystify the initial setup and design choices, providing a structured approach for integrating LangGraph into real-world applications. It addresses the common challenges and decision points faced by teams adopting new AI orchestration frameworks, ensuring a smoother development process. The piece emphasizes the practical considerations for building robust and scalable AI agents. It likely delves into architectural patterns, state management within agentic systems, and how to effectively sequence different AI models or tools into a cohesive workflow. For those focused on production deployment, the guide would cover best practices for reliability, testing, and potential optimizations when scaling AI agents. By offering a "decision-maker's guide," it goes beyond mere syntax, encouraging readers to think critically about the implications of their design choices for long-term maintainability and performance in applied AI contexts. Comment: LangGraph is a critical tool for serious agentic AI development; this guide to building pipelines and making early design decisions is exactly what many developers need to get started right. I Published an A

Engineering managers are almost entirely absent from the AI transformation discourse. There's a structural reason for that, and understanding it is the first step to doing something about it. Engineers write on the internet. C-suite decisions make headlines. Engineering managers absorb pressure from above, complexity from below, and produce outcomes that get credited in both directions. The system doesn't reward the EM voice publicly. But the EM position gives you something that's genuinely hard to replicate: accountability for what happens to the team, combined with proximity to all three layers of the problem at once. That's not a consolation prize. It's a specific kind of leverage, if you decide to use it deliberately. You're accountable for what nobody else fully sees Writers go where the audience is or where the authority sits. EMs are neither, which is why the playbooks keep missing them. Executives get advice that assumes frictionless implementation. Engineers get advice that assumes organizational stability. At the team level, neither holds. The EM isn't the only person with this view. A good Staff or Principal Engineer often has comparable exposure — technical depth, some business context, real influence on architecture decisions. In many organizations, the senior IC has more technical credibility than the EM and less organizational noise to cut through. The difference isn't the view. It's the accountability. When something goes wrong at the team level — delivery slips, quality degrades, an engineer burns out, AI adoption produces incidents instead of velocity — the EM is the one who carries it. That asymmetry is uncomfortable. It's also what makes the EM's perspective structurally different from everyone else's. You don't just see the intersection where the playbooks break down. You're responsible for what happens there. The question isn't whether that position is valuable. It is. The question is whether you're using it actively or just absorbing it quietl

A physiotherapist asked me a simple question a couple of months ago: "Can you build something to run my whole clinic?" So I did — solo, in about a week. Here's the full 2.5-minute walkthrough 👇 What PhysioFlow does PhysioFlow runs an entire physiotherapy clinic from one screen — built for India (₹, GST, WhatsApp, +91): Dashboard — attendance, collections & pending bookings at a glance Patient files — recharge session packs, track usage, auto-generate GST-ready bills Attendance in seconds with a QR scan Online bookings that convert straight into a patient file Reports — daily ledger, revenue, CSV/PDF export Patient portal — patients see their own sessions & prescriptions The stack Next.js + Supabase + TypeScript — multi-tenant, with row-level security so no clinic's data ever leaks to another. Try it Live now with a 14-day free trial, no card needed → https://physioflow.devfrend.com I'm Amar, a full-stack & AI engineer. I design and build products like this end-to-end. I'm open to work — SaaS builds, MCP servers, LLM apps & automation. Reach me on LinkedIn .

This is a submission for the Hermes Agent Challenge , under the Write About Hermes Agent prompt. I've spent the last week testing AI memory failure modes in a public evaluation harness. That work changed how I read agent memory systems. This is a writing submission, not a build submission. I did not build a Hermes Agent project for this challenge. I am writing from the perspective of someone testing how memory failures show up once agents can act. So when I look at Hermes Agent, the question I care about is not only: Can the agent remember useful things? The harder question is: When memory conflicts, which memory is allowed to govern the agent's action? That distinction matters. Hermes Agent is interesting because it is not just a chat interface. Its documentation describes an open-source agentic system with tool use, project context, persistent memory, skills, browser automation, checkpoints, delegation, scheduled tasks, and multiple memory providers. That is exactly the kind of system where memory stops being a convenience feature and starts becoming part of the agent's operating boundary. If an agent can run tools, edit files, browse, delegate work, schedule tasks, and remember across sessions, then memory is no longer just "context." Memory becomes governance. The Memory Problem I Would Watch For In a simple chatbot, bad memory is annoying. In an agent, bad memory can become operational. The failure mode is not only that the agent forgets something. Sometimes the more dangerous failure is that it remembers the wrong thing too confidently. A memory can be: relevant but stale, relevant but low-authority, relevant but superseded, relevant but only context, relevant but not allowed to determine the action. That is the distinction my own tests kept running into. Retrieval systems are usually good at answering: What memory is closest to the user's request? But safety often depends on a different question: What memory is allowed to decide what the agent should do? Thos

1. Why I use Kiro I've been using Kiro for almost 1 year now, I'm using it as a Cloud Architect and also to build side projects for fun. The main reason I use Kiro over other tools is how it works with you as an engineer. Over the months, I've noticed certain patterns in how I use Kiro. Let's go over them: Index 1. Why I use Kiro 2. Pair Programming with Kiro 3. Repeatable workflows as Skills 4. Using Plan, Specs and Agents 5. Council of agents 6. Documentation, Documentation, Documentation Final thoughts 2. Pair Programming with Kiro The most common way that I use Kiro is in Pair Programming. Pair Programming is when there are 2 developers working together on the same task, they can work in tandem or one of them can be the one guiding/planning while the other one does the code. In my case, with Kiro, I'm the one doing the guiding and planning while Kiro is the one executing and implementing the code. I'm also using Kiro as my rubber duck. If I have a new idea or I'm working on a blocking bug, I talk to Kiro so it can give me a different point of view, investigate and steer me into good practices. The main reason for me to do it this way is because once the session is over, I can run a prompt/skill to record everything from the session: Kiro, summarize this session and save it into a .memory folder with the format yyyymmdd and as a markdown So then everything that we've done is going to be recorded there. Do you remember everything that you've done yesterday? Maybe. But what about last week? And what about one month ago? I definitely don't remember it. In the classic Software Development Life Cycle, we have tickets, and we have a way that we can recall all this information, but the more detailed context of why you did it is going to be completely missed. Now, with tools like Kiro, this is possible to remember. You just have a .memory folder where you summarize all your sessions. So in the future, we could have a situation like this: Oh, I don't remember what changes

Environmental activist Erin Brockovich has a new mission.