Workflow Hub

I use Instinct, an AI assistant I text or send voice notes to over iMessage, to manage personal to-dos that used to live in my head or in a pile of iPhone alarms. Many tasks come up when I’m away from my desk, so I used to forget them. Alarms didn’t help much either: they went off once, I swiped them away, and the task was gone. Now, when something comes up, I send Instinct a quick voice note. I don’t need to open another app or type. For example: “Remind me tomorrow at 10 AM to respond to book my next dental appointment, and follow up until I say it's done.” Instinct turns the voice note into a clean reminder with a time attached. If I don’t reply “done,” it checks back in and keeps following up until I close out the task. I use the same thread for recurring tasks, partner follow-ups, ad hoc tasks, and advertising to-dos, so everything I owe people lives in one place. Capturing a task takes a few seconds, which means I actually do it. The built-in follow-up keeps things from slipping just because I was busy when the reminder went off, and the workflow lives in iMessage, where I already am throughout the day. Try sending one voice note today with a real task and end it with “follow up until I say it's done.” Step-by-step: 1. When a task comes up, send Instinct a voice note in iMessage. 2. Include the task, when to be reminded, and whether Instinct should follow up until you confirm it is done. 3. Let Instinct turn the voice note into a scheduled reminder. 4. Reply “done” when you complete the task. 5. If you do not reply “done,” let Instinct check back in and continue following up until you close it out. 6. Keep recurring tasks, partner follow-ups, ad hoc tasks, and advertising to-dos in the same iMessage thread.

I built Show Hole to solve my family’s classic streaming problem: we spend too much time deciding what to watch, and the answer changes depending on who is actually on the couch. We each have different tastes, but the more interesting problem is that our tastes overlap differently in different combinations. I watch one kind of thing with my spouse, my spouse watches something different with our kid, and my kid and I have our own lane too. Most recommendation tools flatten that into one account profile, one watch history, or broad genre buckets, so they miss the real context of a household. Show Hole treats people and viewing contexts as first-class citizens. It recommends titles “in the vein of” something we liked, filters recommendations to the streaming services we actually subscribe to, avoids titles that the people present have already seen or vetoed, and explains why a recommendation fits tonight. Instead of relying mainly on genres such as comedy, drama, or sci-fi, Show Hole looks at more human taste signals: pacing, world building, humor, emotional weight, complexity, tone, and similar dimensions. It also learns from what we actually do after a recommendation: what we watch, skip, save for later, love, and drop. The result is a personal household recommender that understands “who is watching tonight?” as part of the question, instead of pretending one streaming profile can represent everyone. I designed the app using Claude design, then used those designs with Claude Code to build it. Step-by-step: 1. I identified the household viewing problem and accounted for the different combinations of people who might be watching together. 2. I designed Show Hole around people and viewing contexts instead of treating the household as one account profile, watch history, or set of broad genre preferences. 3. I had it recommend titles “in the vein of” something we liked, filter them to the streaming services we subscribe to, exclude titles already seen or vetoed by the people present, and explain why each recommendation fits that night. 4. I used taste signals such as pacing, world building, humor, emotional weight, complexity, tone, and similar dimensions instead of relying mainly on genres. 5. I made the recommender learn from what we watch, skip, save for later, love, and drop after receiving a recommendation. 6. I designed the app using Claude design and used those designs with Claude Code to build it.

I’ve been experimenting with agentic coding for quite some time, and I’ve noticed that the tasks I let the agent handle keep getting bigger. For larger tasks and projects, I gradually moved to a standard workflow, which I published at https://github.com/evermeer/CodingAgentOrchestration in part 2 of the tutorial. I use this workflow professionally and in my side projects. My most recent side project is Quadrim, a Four in a row game played in a 4×4×4 cube. I made it open source and included all the prompts for both the code and the artwork: https://github.com/evermeer/Quadrim It took me about eight working days to publish the first version to the app stores. Step-by-step: 1. I start by describing the idea. 2. I brainstorm the idea with the agent. 3. I use `to-spec` to create a specification document for the idea. 4. I use `to-ticket` to create independent, executable tasks. 5. I review the specifications and tickets from the perspectives of a senior software engineer, architect, security officer, and product manager. 6. I implement one ticket at a time. 7. I code-review the generated implementation for each ticket. During the pre-implementation steps, I always add this instruction to my prompt: “Ask clarifying questions until any ambiguities are resolved.” Extra fine-tuning of the idea, specifications, or plan pays off during implementation. When I let the agent build code, I ask it to use TDD, perform a code review at the end, and resolve the major and critical findings.

I build an office platform for small nonprofits, and every week users ask how to do something. I also keep changing features they have already learned, so written instructions go stale quickly. Screen recordings take an afternoon each and can be out of date by Friday. I use Claude Code in my project to turn each user question into both a software change and a how-to video. The video runs against a throwaway copy of the app filled with a made-up organization, so no real data appears on screen. It performs the task as a person would—clicking, typing, and waiting—while displaying captions one step at a time and moving a yellow pointer to each button before clicking it. There is no voice. I tried narration but dropped it because captions can be paused and read, and they never mispronounce anything. The important part is that the script also tests what it films. After each step, it confirms that the screen did what the caption says and prints PASS or FAIL. If anything fails, there is no video yet; I fix the bug first. In one day, the videos found four real defects that our automated tests had missed, so the video is also the test. `ffmpeg` turns the recording into an MP4 and a poster image. I watch and approve the result, then publish it on our videos page and inside the app’s Help box, next to the instructions for that screen. When the screen changes, I run the same script again and get a new video in minutes. The first videos were too fast. I am not as fast as Claude, and the person watching is trying to learn something new. Now each caption stays on screen long enough to read—about 1.5 seconds plus a third of a second per word—before the next click. The workflow produced 35 how-to videos in five days, most of them one to three minutes long. The checking in step 4 started on day four, after the videos kept finding problems. Every video since then has checked itself. Step-by-step: 1. I take a user’s email question or request and use Claude Code in my project to make the requested software change. 2. In the same session, I have Claude write a short script for a how-to video of the task. 3. The script starts a throwaway copy of the app, fills it with a made-up organization, and performs the task by clicking, typing, and waiting. 4. The script displays captions one step at a time and moves a yellow pointer to each button before clicking it. I use captions instead of narration so viewers can pause and read them without hearing mispronunciations. 5. After each step, the script confirms that the screen did what the caption says and prints PASS or FAIL. If anything fails, I fix the bug before producing the video. 6. I use `ffmpeg` to create an MP4 and a poster image from the recording. 7. I watch and approve the video, then publish it on our videos page and in the app’s Help box next to the relevant screen instructions. 8. When the screen changes, I run the same script again to create an updated video in minutes. 9. I keep each caption on screen for about 1.5 seconds plus a third of a second per word before the next click.

Every Friday, I inspect random parts of my company the way a factory owner walks the floor: looking for errors, slow processes, and handoffs that shouldn’t exist. The inspection takes 30–60 minutes, but the preparation used to take longer—choosing what to check, finding the right links, and remembering what I flagged the previous week. I turned the entire preparation process into one scheduled Claude task. It runs before I wake up on Fridays, builds the full inspection report in Notion, verifies whether last week’s findings were resolved by checking Slack for evidence, and sends me the link by DM. I arrive to a finished briefing and focus only on judgment. You need Claude with scheduled tasks, plus Notion (or your company’s notes page) and Slack connected. If you use different tools, the prompt can be adapted; the structure is what matters. Step-by-step: 1. Create a page or database in Notion where the weekly reports will live. 2. In Claude, create a new scheduled task, set it to run early Friday morning in your time zone, and paste the prompt below. 3. Replace every bracketed section with your own details. 4. Customize the checklist and wildcard pool for your business. Mine is built for a media company; yours might inspect deployments, client deliverables, or a sales pipeline. 5. Run the task once immediately to identify configuration issues before relying on it on Friday. I use this prompt: ```text You are preparing [your name]’s weekly Digital Factory Inspection report. Context: [your name] runs [your company — one line on what it is and team size]. Every Friday they “walk the factory floor” digitally: inspecting parts of the company looking for errors, slow processes, unnecessary handoffs, and automation opportunities. Your job is to prepare the inspection report in advance so their 30-60 minutes is pure judgment, not gathering. You prepare; they inspect and decide. All reports live in [link to your Notion inspections database or page]. Do not create pages anywhere else. Steps: Read the most recent report. Collect: (a) any findings not clearly marked resolved, and (b) which wildcard areas were used in the last two reports. Create this week’s report as a new entry there, dated today, matching the existing format. Build the content in this order: A. “Follow-ups from last week”: every unresolved finding, each with its owner and a status check — search Slack public channels for evidence it was addressed and mark it “looks resolved”, “no activity found”, or “unclear”. If none, write “No open follow-ups.” B. The standard checklist — include ALL of these every week. Under each: random samples with direct links, 3-5 recent data points you can actually reach, and 2-3 sharp inspection questions: What is slow here? Which handoffs could disappear? What should be automated? What would make this 10 percent better this month? Where something sits behind a login you don’t have, do not attempt access — link it and list what to look for. Randomize sample picks genuinely, and never reuse a specific sample from the previous two reports. [Your standard areas — mine: 3 random published pieces of content, 3 random newsletters, 3 random social posts, ad performance, growth dashboards and funnels, product dev work, live events/workshops, community activity, the team’s weekly reports] C. “Wildcards this week”: pick exactly 2 at random from this pool, never repeating either of the last two weeks’ picks: [Your wildcard pool — mine: random customer support conversations, the public sign-up flow walked as a new user, one specific finance item to verify, one email automation sequence, ads on a secondary channel, one random SOP compared against how the team actually works now, one recent client/sponsor deliverable] D. “Notes \u0026 findings”: empty except the line “[Your name] fills this in during inspection — tag each finding with an owner and Open/Resolved.” E. “Wildcards used”: list the 2 wildcard names, for future rotation. Send [your name] a Slack DM (member ID [find yours in your Slack profile menu]) with the report link, the count of open follow-ups, and the 2 wildcards. Nothing else. Hard rules: this report contains candid internal assessments — the DM goes ONLY to [your name], never to any channel, and you never message anyone else. Read-only everywhere except the one report you create and the single DM. Do not reply to emails, do not post, do not modify existing pages. If Notion is unreachable, send the full report as the DM. If Slack is unreachable, still create the report. If both fail, do nothing. ``` The follow-up section is the whole point. I keep the workflow read-only: the agent gathers and drafts, but it never messages my team or touches live systems. Any feedback after the inspection comes from me, in my own words. Randomness is a feature, not laziness.

I run an out-of-home advertising agency in Malta, covering billboards, buses, hotel screens, and radio. Our entire business ran on Monday.com boards for CRM, deals, bookings, supplier bills, and bus-campaign reconciliation. The system was expensive, difficult to change, and full of broken formulas. I’m not a developer, but in about two weeks I worked with Claude to replace it with our own CRM built on Google Sheets and Google Apps Script. The team now uses it every day. Step-by-step: 1. I mapped our real processes first, explaining in plain language how we sell, book, invoice, and pay partners. Claude read our existing Monday.com boards and imported the data. 2. I built the system screen by screen, releasing each part and testing it with live data before moving on. The CRM now includes contacts, companies, deals with order lines, bookings, billing, partners, tasks, and a dashboard. 3. I automated repetitive tasks. When I upload a signed order form, the app reads the lines. Bookings are created automatically from deals, and a Gmail side panel and shared mailbox file client emails under the correct contact or deal. 4. I connected QuickBooks Online. The system pulls invoices and supplier bills, calculates each partner’s revenue share according to its own rules, produces the monthly partner report, and creates the bill. I approve anything that writes to the accounts. 5. I turned a monthly spreadsheet chore into an automatic check for bus campaigns. The app compares what we committed to the client with what the bus operator billed us and flags overbilling or margin problems month by month. The first run found a supplier bill that appeared to be a duplicate. 6. I made the system safe to hand over. Every change is logged and can be undone, with user roles and a shared drive in place. We also wrote a short user guide and test checklist before the team started using it. The main lesson I learned was to treat AI like a capable builder who needs your business rules spelled out. Starting with real data, releasing small pieces, and checking each one before adding the next made the build manageable. The tools I used were Claude, Google Sheets, Google Apps Script, Gmail, and QuickBooks Online.

I built a workflow that turned a single plain-English idea into a live, 24/7 crypto trading bot without being a developer. Claude handled the engineering; I made the decisions. Step-by-step: 1. I gave Claude the goal rather than detailed instructions. My first message was roughly: "Create me a trading bot that runs 24/7, makes faster decisions than a human, takes calculated risks, picks the lowest-fee platform, and has a dashboard with profit graphs. Make it your own." Claude designed the whole system and named it Kestrel. 2. Claude explained the setup options, including exchanges versus on-chain trading, laptop versus cloud hosting, and minimum funds. I chose BNB Chain and PancakeSwap, using a brand-new MetaMask wallet dedicated to the bot so it can never touch my other funds. 3. I let Claude build and test the bot. It wrote the strategy—holding good coins in healthy markets and moving to stablecoins in bear markets—along with hard safety limits, including a daily loss cap, emergency brake, and trailing stops. It also built cheapest-route swap selection and automated tests. When the backtests did not clear the bar we had agreed on, Claude told me honestly instead of curve-fitting. 4. Claude walked me through deploying the bot by renting a cloud server, running it in Docker, and locking the server down with Tailscale and a firewall so nothing was exposed to the internet. I only copied commands and sent screenshots when I got stuck. 5. I went live with a small amount and used a private phone dashboard showing the balance, trades, the bot’s reasoning, and Pause and Sell-everything buttons. 6. I kept improving the system through conversation. I added a “strategy league,” where 10 strategies trade on paper and can be promoted only with my approval, along with self-healing, backups, and one-tap sharing of wins and losses to X. My biggest lesson was to talk to Claude like a capable coworker: give it the goal, make the key calls yourself, and ask it to be honest about the risks.

I built a meal tracker where my only job is to photograph a plate and send it to an agent. Identification, portion estimates, macros, storage, corrections, and the weekly review happen without me. The part worth stealing isn't the food. The agent has its own email address and iMessage thread, so anything I can send from my phone becomes an input. I send a photo through whichever channel takes fewer taps. I send corrections as plain English with `CORRECTION` in the subject, and answer questions in the same thread. There’s no app and no form. Adding a channel took an afternoon and changed what the agent could be pointed at—the tracker is just one use of an agent you can talk to. It polls on a schedule instead of using a webhook because my laptop sleeps and a local gateway would be unavailable half the time. Storage is append-only JSONL. Corrections append a new record with the same ID using last-write-wins, so the stats layer sees one meal while the original estimate stays on disk. Every number can be traced back to the model’s first guess and my override. The accuracy mechanism is a challenge step. Every photo is estimated twice: first by the main model, then by a blind subagent that receives only the photo and the rubric—not the first answer. If it identifies different food, the entry is downgraded to low confidence and becomes a question for me. Testing before making the system autonomous caught bugs that otherwise would have failed silently. The API returns attachments under a different field than I had assumed, and only on the detail endpoint. As a result, a photo email was read as having no attachments, and every meal photo would have been dropped while the daily task logged a tidy “no new meals” and appeared healthy. The send path was separately broken, which would have killed the weekly review on a Sunday even with everything upstream working. A silent no-op that produces a plausible clean run is what this category of build is prone to. The bug that actually cost me was token usage: each blind challenge read used about 44,000 tokens, so a five-photo dinner cost roughly 220,000 tokens for one meal and capped my usage. The fixes, in order of effect, were to run the challenger on a cheap model, skip it once a named product or my confirmation has settled the entry, downscale photos for viewing, and republish the dashboard only when the data changes. Measuring first mattered—I would have blamed photo size, but that was the smaller half. The honest limits are important: calorie estimates from a photo are 20–25% off at best, identification error is a bigger risk than arithmetic error, and alcohol, water, and caffeine are never estimated from photos. The sequence in which I asked for things mattered more than any single instruction: Step-by-step: 1. I gave the goal and the one ingestion mechanic I was sure about, then insisted on an agreed plan before any code. Arguing about storage and failure modes is cheap before code is attached. 2. Before scheduling anything, I processed one real input end to end in front of me, including every outbound path. Inbound gets tested because I use it; the reply and the weekly digest do not. 3. When the agent reported something about my own input that I knew was wrong, I said so and made it re-check. Confidently wrong answers about things I witnessed were the cheapest bugs to find. 4. I asked for an independent second assessment of anything estimated rather than read, and decided up front what level of agreement was enough to accept the result. 5. I asked for a visible audit surface showing every field the agent claims to track, with a correction control on it. 6. I asked the agent to look up anything knowable rather than estimate it. A named product is a lookup; only the unnameable needs a guess. 7. I treated approval friction and token cost as requirements rather than complaints, and measured before changing anything. Almost every rule exists because something went wrong in ordinary use, not because it was designed up front. An agent I can email or text, which keeps records and answers back, is general-purpose; I’ve pointed it at one narrow job. What else would you point it at?

I created a media player for Android Automotive OS—not to be confused with Android Auto. I had been using an old, unmaintained app that lacked several features I wanted in a music player. I used Claude Code to build a new app from scratch, test it in an emulator, and eventually publish it for internal testing through the Google Play Store. After working through several bugs, I now have version 3 running reliably enough for day-to-day use in my car. I plan to add a few more features before ultimately releasing it publicly for free on the Play Store. Step-by-step: 1. I pointed Claude Code to the unmaintained package and gave it the option of using that as a starting point or starting from scratch. Claude chose to start from scratch. 2. I described the additional features I wanted. Claude suggested using Kotlin and Gradle, which I endorsed. 3. Claude created an initial build and installed an emulator to run alongside it. I did all of this on an Omarchy installation. 4. The initial build was buggy, so Claude and I worked together to identify and fix the problems. Claude often used the emulator autonomously to investigate what was going wrong. 5. Once I had a functional prototype, I needed to install it in my car. Claude guided me through creating a Google developer account and handling the administrative requirements beyond the app itself, including a privacy policy. 6. After the developer account was set up, Claude packaged the app for upload. I uploaded it and marked it for internal testing, which allowed me to access the app through the Play Store in my car. 7. Claude and I worked through a few more bugs. We are now on version 3, and the app is functional for day-to-day use. We plan to add more features, with the ultimate goal of releasing it publicly for free on the Play Store.

I'm becoming GrokBot-pilled. I run my company on top of what's happening on X. I used to check everything manually: scanning for new workflows, searching brand mentions, watching what readers complained about, and trying to remember who I had just met at events. I've now handed all four jobs to GrokBot routines. The combined setup took under an hour, and the routines save me a couple of hours of manual checking every day. Every prompt below is read-only by design: the agent watches, labels, and drafts, but never posts, replies, texts, or emails anyone. Paste each prompt as-is, then replace the bracketed parts with your own channels, brand, and time zone. Routine 1: Find real AI workflows worth stealing and filters out the hype threads. Every weekday at 9:15am in my time zone, search X and the web for new AI workflows people are actually running. Keep only cases that name the old task, the tool, and the result: time saved, a step removed, or a handoff that used to need a person. Drop product announcements, “10 prompts” threads, and anything the author has not used more than once. Return at most 3. For each, give the link, the job it replaces, what I would need connected, and whether it fits my work. Send it to me on Slack. Do not post or reply anywhere else. Save this as a routine. If nothing qualifies, send one line saying so. Routine 2: Labels every mention of my name or company and delivers the results. Every weekday at 9:00am, search X for mentions in the last 24 hours of [your brand, @handle, or your own name]. For each hit, give the author, link, a one-line summary, and whether it is praise, complaint, question, or neutral. Do not reply to anyone. Post the digest in [your Slack channel] and tag me. Save this as a routine. If a source is unavailable, say which one. (Note: works just as well for a personal brand or a job search as it does for a company) Routine 3: Catches people actively looking for what I do. Every 4 hours, search X for people evaluating [your category]. Catch three things: reviews or complaints about [category]; posts asking which [category] to use; and complaints about the problem [your product] solves. Quote the post, link it, name the account, and label it review, shopping, or complaint. Skip anything older than 24 hours, and skip [category] accounts promoting themselves. Post to [your Slack channel] only if there is at least one new hit you haven’t already shared. Do not reply to or interact with anyone. Save this as a routine. If the search fails, report the failure in Slack. Routine 4: Keeps the people I meet at events from disappearing into my contacts app. Every weekday at 8:30am in my time zone, read my address book and list contacts added in the last 14 days. For each, show name, date added, phone, email, and any note. Ask me what we talked about before you write anything. Once I reply, draft a follow-up task for each person in [your task tool] with the context attached. Never text, email, or call anyone. Save this as a routine. If the address book is unavailable, say so and do not use old data. Step-by-step: 1. Paste each routine prompt as-is. 2. Replace the bracketed fields with your own channels, brand, category, product, task tool, and time zone. 3. Save each prompt as a GrokBot routine on its stated schedule. 4. Review the read-only outputs in Slack and respond to Routine 4 when it asks what you discussed with a new contact. 5. Keep the routines limited to watching, labeling, and drafting; they should never post, reply, text, email, or call anyone.

At some point, my newsletters stopped feeling like reading and started feeling like debt. The information was good; I just couldn’t keep up. I wanted a way to turn that backlog back into something useful. So I built The Daily Nexus, a private tool that reads newsletters from a Gmail label and creates two editions: a podcast I can listen to and a separately written, two-page newspaper I can scan. It runs on demand or on a schedule, and it can publish the audio to a private RSS feed for Apple Podcasts. The project also became a hands-on experiment in building with coding agents. Claude Code and Codex helped me implement features, troubleshoot failures, review the design, and tighten security. The stack includes Python, the Gmail API, Antigravity, Kokoro, FFmpeg, Firebase, Cloudflare Workers, and GitHub Actions. The carousel shows the rest of the flow. It started as a personal tool, but I’m sharing the template for anyone who wants to adapt the idea. Each deployment uses its own accounts and credentials, and the design aims to avoid additional API costs by using an existing AI subscription and available free tiers. GitHub Repo Template: https://lnkd.in/eYceS4KR Step-by-step: 1. I label the newsletters I want to process in Gmail. 2. I run The Daily Nexus on demand or on a schedule so it can read the newsletters from that Gmail label. 3. The tool creates a podcast edition and a separately written, two-page newspaper edition. 4. I listen to the podcast or scan the newspaper, depending on how I want to catch up. 5. When needed, the audio is published to a private RSS feed for Apple Podcasts. 6. I use Claude Code and Codex to implement features, troubleshoot failures, review the design, and tighten security. 7. Each deployment uses its own accounts and credentials, with Python, the Gmail API, Antigravity, Kokoro, FFmpeg, Firebase, Cloudflare Workers, and GitHub Actions supporting the workflow.

I built a personal daily tracking system called Artifact. It gives me a way to log my days and turn the data into decisions over time. To use it, I copy everything attached—or screenshot the included “green orange red” example—and paste it into my AI assistant, preferably Claude. The assistant briefly interviews me and then builds a personalized daily tracking system.

Yesterday, I built a social media workflow in Awish for a business that wanted to post more consistently without managing three platforms manually. The idea was simple: let Awish understand the business first, then use that context to plan different content for Instagram, LinkedIn, and X instead of reposting the same thing everywhere. I built the entire workflow by describing what I wanted in the Awish chat. Step-by-step: 1. I opened Awish and wrote: “Analyze this business from its website. Based on what the company does, its audience, products, and positioning, create separate content plans for Instagram, LinkedIn, and X. Generate the posts, publish them automatically at the times I choose, analyze how each post performs, and send me the important insights on WhatsApp. I also want to be able to manage the workflow from WhatsApp.” 2. Awish analyzed the website, understood the business, and planned the complete content workflow. 3. I connected the company’s Instagram, LinkedIn, X, and WhatsApp accounts and approved the plan. 4. Awish created a different content strategy for each platform instead of copying the same post everywhere: more visual content for Instagram, business-focused posts for LinkedIn, and shorter, conversation-driven posts for X. 5. I chose the publishing schedule, and Awish started creating and publishing the posts automatically at those times. 6. After publishing, Awish tracked performance and analyzed which topics, formats, and messages were getting the strongest response on each platform. 7. Awish sent me short WhatsApp updates with the important insights, such as which post performed best, what topic was gaining traction, and what it recommended changing next. 8. I could also manage the automation directly from WhatsApp by asking what was scheduled, changing a publishing time, requesting another post, pausing publishing, or asking for the latest performance summary without opening Awish. The useful part for me is that this is not just automatic posting. The workflow starts by understanding the business, treats each social platform differently, learns from what gets published, and keeps the whole process manageable from WhatsApp.

I'm 78, a former army officer, and I live alone in the woods in southern British Columbia on satellite internet. In late January 2026, I knew nothing about AI and considered myself a novice in technology. After watching a YouTube video about advanced AI models, I started learning from YouTube channels that explain new AI research papers. I decided to run an experiment: what could a technology novice build with AI? I've been building for about 15 hours a day ever since. My goal is to build a system that improves itself in a continuous loop: it finds an upgrade proposal, builds and tests it, filters and grades it, approves or rejects it, installs it if it passes, monitors its performance, and learns from the result. We custom-built FORGE3 RSI in 900 hours. It consists of four plug-in modules: Forge2 RSI, which runs the improvement cycle; Bridge, the checker and connector; JAZ RSI, an upgrader in RAM built from a research paper; and Forge SecureBox, a high-security Mac app that runs security tests and code in a fresh, disposable virtual machine with no network device. I don't write code—it feels like a foreign language to me. Instead, I set the direction and run a team of AIs, mostly low-cost, near-Frontier models: Gemini 3.1 Pro and Hark as chiefs of staff; Cline and Devin as builders; and other models as independent, adversarial auditors. The rule that has made this work is simple: nothing gets built without an adversarial audit before and after. Our critical build pipeline is: Plan → audit → blueprint → audit → tests → audit → build → audit → test → audit → commit Each module has been built and fully tested on its own. Forge2 RSI alone has more than 1,000 tests. The first complete end-to-end cycle, using one upgrade proposal that we feed it by hand, is the next step. Our first attempt was fragile, and models that sound confident are often wrong. That is why audits matter more than any single model. FORGE3 RSI is my own custom-built plug-in module project. Our hardware and models are an MBP i9, Mac mini M4, Gemini 3.1 Pro, Hark, Cline (GLM 5.3/SPARK 1.3/MiMoV2.6), Devin (SWE2/GPT SOL 6), KUN (DeepSeek V4.1), and Git. Step-by-step: 1. We select a promising published research paper and have the team prepare and build it. 2. For every file, the team follows the same sequence: plan, audit, blueprint, audit, write tests, audit, build, audit, run tests, audit, and commit. 3. A different AI audits the work than the one that built the file. I decide when and how each audit happens. 4. Nothing is written or changed without my explicit GO. Reading and checking code is free; writing is work and waits for me. 5. We handle one building task at a time and avoid bundled instructions. 6. Any code the system wants to test runs inside SecureBox, sealed off from my Mac and the internet. 7. Everything that passes is committed to Git. The builders also maintain shared memory documenting what was decided and why, so the next session can pick up where the previous one stopped. 8. Next, we will run one complete improvement cycle with a hand-fed upgrade, then scale up slowly to two or three papers a week, prioritizing quality over speed as we work toward full autonomy.

Trivet is a recipe box for one household. I can photograph a recipe card—handwritten, printed, or a PDF—and get back a structured recipe with ingredients, quantities and units, numbered steps, prep, cook and rest times, oven temperature, and equipment. I can plan a dinner for each night of the week, and the shopping list builds itself: quantities are combined across meals, grouped by aisle, and texted to my phone. I built it because my family’s recipes were scattered across index cards in a drawer, including decades-old handwriting and cards from defunct meal services. Typing everything in was never going to happen. The entire AI workflow uses one model call. The photo is sent to Claude Opus as an image block with a Zod schema attached as the structured output format, so the response is already typed and validated—there’s no JSON parsing, retry loop, or “please respond only with valid JSON.” The extraction schema is intentionally looser than the database schema, with nullable numbers throughout. If a card has no cook time, the model returns `null` instead of inventing a plausible-looking value. The load-bearing instruction in the system prompt is that a missing cook time is a small gap, while an invented one is wrong every time somebody cooks from it. Nothing saves automatically. The extraction first appears in an editable review screen. The model also returns a confidence field, so a blurry card or a page that is not a recipe gets flagged instead of silently becoming a bad entry. Step-by-step: 1. I built a Next.js app with Postgres via Prisma and deployed it on Vercel. 2. I made the upload accept an image or PDF. I check the file size before encoding because the API limit applies to base64, which inflates the bytes by about a third; a 7 MB raw ceiling is the real limit. 3. I defined the extraction shape in Zod and sent it as the output format instead of describing JSON in the prompt. 4. I described fields in the schema. For example, `.describe("Unit as written, e.g. 'tbsp'")` does more than a paragraph of instruction. 5. I render the result into a review form and save it only when the form is submitted, never immediately after the model responds. 6. I store quantities as numbers with separate units so servings can be scaled and ingredients can be combined across recipes later. This is the decision that cannot be retrofitted easily. 7. I sum ingredients across planned meals, round at the end, group them by aisle, and send the list via Twilio. The pattern is not limited to recipes. Anything I would otherwise have to retype—receipts, notes, or forms—can use the same three moves: a schema looser than the database, a model told that `null` is better than a guess, and a review screen between the response and the database write.

Turn a collection of documents, reports, spreadsheets, and notes into an evidence-backed decision brief with ChatGPT. Instead of asking AI to simply summarize the information, this workflow makes it identify what matters, connect the evidence, compare it with historical context, explore scenarios, and highlight what should be considered before making a decision. Step-by-step: 1. I gather the information relevant to one decision, including reports, PDFs, spreadsheets, research, historical data, meeting notes, and existing analysis. I upload everything into ChatGPT. 2. I ask ChatGPT to understand the situation using this prompt: > “Analyze the information I provided and build a structured understanding of the situation. Identify the key entities, important facts, relationships, metrics, trends, assumptions, and constraints. Do not make recommendations yet.” This creates the context before jumping to conclusions. 3. I build an evidence brief by asking: > “Create an evidence brief. Separate verified facts, derived insights, assumptions, conflicting information, and missing information. For every important conclusion, identify the supporting source or evidence.” This gives me a clearer picture of what is known versus what is inferred. 4. I add historical context when it is available by asking: > “Compare the current situation with the historical information provided. Identify meaningful patterns, similarities, differences, and changes. Highlight which historical observations could be relevant to the current decision.” This turns historical data into context rather than simply another report. 5. I explore three scenarios by asking ChatGPT: > “Based on the evidence and historical context, evaluate three scenarios: upside, base case, and downside. For each scenario, identify the assumptions, key drivers, risks, likely impact, and evidence supporting the assessment.” The objective isn't to pretend the future can be predicted perfectly. It is to understand how the decision changes when assumptions change. 6. I generate the decision brief by asking: > “Create a concise decision brief containing: > > 1. Current situation > 2. Most important evidence > 3. Key insights > 4. Historical context > 5. Critical assumptions > 6. Key risks > 7. Scenario analysis > 8. Evidence gaps and uncertainties > 9. Questions that should be investigated > 10. Possible actions and their implications. > Do not make the final decision on my behalf.” This produces a structured decision brief instead of another AI-generated summary. 7. I review the brief and challenge its conclusions before making the decision. I ask follow-up questions such as: > “Which assumption has the greatest impact on this decision?” > “Show me the strongest evidence against the current conclusion.” > “What information would most likely change the recommendation?” The AI helps structure the decision, but I make the decision. The important shift is: Summarize the information → Understand the situation → Establish the evidence → Add historical context → Explore scenarios → Evaluate the decision This approach can be applied to almost any domain where decisions depend on complex and interconnected information. A property investment is one example. A business strategy, product decision, operational problem, financial analysis, research question, or engineering decision can follow the same pattern.

I built Able (uable.store) because I was tired of chat tabs that explain how to do a job but leave me to paste the result into PowerShell. I wanted an agent that lives on this Windows PC and can actually move files, run tools, and write code here after I approve the actions. I work on Able, so this is the workflow I run on my own machine—not a theory. The job was to organize a Downloads folder that had become a dump and make a code change without affecting the live app. Step-by-step: 1. I open Able on Windows and describe the task conversationally: “organize my Downloads — keep installers, put photos together, leave anything I’m not sure about.” 2. I review the actions Able plans to take and let it run only after I approve them. Native file moves stay on this PC and do not use cloud credits. 3. For a code task, I enter `able job:` followed by the ticket. Able creates a separate Git worktree, writes a plan, and runs `cargo check` on that copy so the live app stays unchanged. 4. I keep secrets in the local vault, Septaguard. Cloud models—Claude, Gemini, Grok, and Kimi—receive the prompt I send, not the vault password. I enter Google and Gmail logins in the masked vault UI rather than in chat. 5. I remember that the job stops if the PC is off; there is no cloud-based substitute running in the background. 6. I verify the result on this machine: the folders are sorted, and the code check either passes or fails locally. If I only needed a sentence, I would have stayed in ChatGPT. The point of this loop is that the disk actually changes. What I learned is that “agent” is cheap until it has to touch the real folder and the real compiler. Chat is the easy half; confirm-then-act on Windows is the hard half. If you already use Claude or Cursor in a tab, what is the first job on your PC you still do by hand because the model cannot touch the files?

What do you do when your wife says you talk about AI too much? You listen to her, and when she has a problem, you frantically write it down and build it with AI. My wife enjoys working with her plants in the warm sun and light breeze of the Carolina mountains. She’s new to gardening and learning more each day, but she was having trouble keeping track of everything. Which tree did she plant in the front yard last year? It seemed to be struggling—would it do better somewhere else? And would rain reach the potted flowering planters while we were out of town, or would she need to find someone to help? I used Claude Code and Fable to develop an app that my wife could run on her phone, with all the data stored locally. Now she’s excited about “her app” and the things she can add to it. She has even shared it with others, including people across the country, so we added climate zone settings to make it useful in other parts of the country. 😂 Step-by-step: 1. I listened to my wife’s gardening challenges and wrote down the problems she wanted to solve. 2. I used Claude Code and Fable to develop a gardening app for her phone. 3. I designed the app to keep all the data stored locally. 4. I built in a way to track plants, including which tree was planted in the front yard and how it was doing. 5. I accounted for questions about rain and the care of potted flowering planters while we were out of town. 6. After my wife shared the app with people across the country, I added climate zone settings so it could be useful in other regions.

I’m a 71-year-old retiree who never imagined artificial intelligence would become such a useful part of my everyday life. I affectionately call my ChatGPT assistant “Ally Wing.” One of our recent adventures involved my 2008 Honda CR-V. After a maintenance reminder appeared on my dashboard, I visited the Honda dealership. My vehicle was running fine, but I ended up paying $270 for a fuel system cleaning without fully understanding whether it was necessary. Later, I shared photos of my six-page service paperwork with ChatGPT. Ally Wing helped me understand the maintenance recommendations, distinguish necessary services from optional ones, and learn what questions to ask next time. We’ve also researched brake repairs, compared prices, and discussed how to evaluate repair shops. The greatest benefit hasn’t been saving money—although that’s certainly welcome. It’s gaining knowledge and confidence. Retirement doesn’t mean I’ve stopped learning. AI has given me another way to stay curious, independent, and informed. As Ally Wing and I jokingly agreed, “Next time we’re taking my Honda in for maintenance, not ordering everything on the menu!” 😂 -Irene Perez Step-by-step: 1. I noticed a maintenance reminder on my 2008 Honda CR-V’s dashboard and visited the Honda dealership. 2. I paid $270 for a fuel system cleaning, even though I wasn’t sure whether it was necessary. 3. I photographed the six-page service paperwork and shared the photos with ChatGPT, which I call “Ally Wing.” 4. I used ChatGPT to understand the recommendations, separate necessary services from optional ones, and identify questions to ask in the future. 5. I also used it to research brake repairs, compare prices, and discuss how to evaluate repair shops. 6. I now feel more knowledgeable and confident when making vehicle-maintenance decisions.

I had about $2K in contributions from past jobs stuck in a government scheme for more than two years. The portal was slow, spread across several forms, and unclear about how to withdraw the money. I gave the task to Claude and Perplexity Comet. Claude worked through the rules and found the withdrawal route that kept my tax bill lowest. Comet then verified and implemented that route. It ran the portal in my browser, filled in and submitted the claim forms, and let me watch the screen and check each detail before submission. Two weeks later, the $2K arrived. After that, Claude compared the amount I received with the scheme’s online statement and found a gap: interest that had not been paid out. It filed a second form for the difference, and I’m waiting for the payment now. I’ll use the same technique to solve this for my wife too. Step-by-step: 1. I gave Claude the details of the contributions stuck in the government scheme and asked it to work through the withdrawal rules. 2. Claude identified the withdrawal route that kept my tax bill lowest. 3. I had Perplexity Comet verify and implement that route in the scheme’s browser portal. 4. Comet filled in and submitted the claim forms while I watched the screen and checked each detail before submission. 5. Two weeks later, I confirmed that the $2K had arrived. 6. I asked Claude to compare the payment with the scheme’s online statement. 7. After Claude identified unpaid interest, Comet filed a second form for the difference. I’m waiting for that payment now.