Workflow Hub

I built a self-hosted wine cellar app that runs in a Docker container on my home NAS and doubles as a personal AI sommelier. I add a bottle by taking a photo of its label with my iPhone. A vision model identifies the wine, vintage, region, and grape varieties, while a second model estimates its market value and writes tasting and pairing notes. The app renders a visual map of my rack, so I can see exactly where each bottle is and pull one without hunting. My ratings feed a taste profile built from my own history, which the AI sommelier chat uses to answer questions such as what to open with dinner or which bottles are drinking at their peak. The feature I use most is Scan & Check. When I’m in a store, I photograph a bottle and get a Collection Fit score based on my profile before buying it. The app also exposes an MCP endpoint, so Claude can search my cellar, score a wine, or list my top-rated bottles from any conversation. I built it almost entirely with Claude Code and use it daily from my phone. Step-by-step: 1. I run the wine cellar app in a Docker container on my home NAS. 2. I photograph a bottle’s label with my iPhone when adding it to the collection. 3. A vision model identifies the wine, vintage, region, and grapes, and a second model estimates market value and generates tasting and pairing notes. 4. I use the app’s visual rack map to locate bottles by slot. 5. I rate wines so the app can build a taste profile from my history. 6. I use the AI sommelier chat to choose what to open, including bottles that are drinking at their peak. 7. I use Scan & Check in stores to photograph potential purchases and review their Collection Fit score. 8. I connect to the app’s MCP endpoint so Claude can search my cellar, score wines, and list my top-rated bottles from a conversation.

I’m building a mobile game called ExoLab Blast with Astra. I developed it through conversation, starting with the basic concept and continuing through testing the game mechanics. Astra also created and repeatedly updated the game’s graphics and UI based on my feedback. Step-by-step: 1. I discussed the basic concept for ExoLab Blast with Astra. 2. I used Astra to build out the mobile game. 3. I tested the game mechanics. 4. I gave feedback on the graphics and UI. 5. Astra created and updated the graphics and UI multiple times based on that feedback.

I'm not a professional developer, but I built Sujido, a live word puzzle, with multiple AI agents. For months, they gave me bloat, lost context, and slop, and neither I nor the models could untangle what had been decided or why. I eventually settled on two agents: one Claude instance that plans and reviews, and another that builds. I imposed rules on both, one failure at a time. None of these rules would surprise a professional engineering team, but nobody handed them to me. That's the key lesson I learned: agents won't bring the discipline for you. You have to impose it. When I showed the first game revision to strangers, their verdict was that the first half was "boring" and the second half was novel. It was hard news to hear about my baby, but I took the feedback to heart and cut the game in half. That required surgery on code tangled across both halves, followed by a rebuild under the rules below. Before the rebuild, a separate AI audit graded the codebase B−. Afterward, it graded it A−. The deployed site shrank by 97.5%, and every serious finding was fixed and locked in by an automated test. The models didn't change drastically, but my operating rules did. Step-by-step: 1. I wrote a constitution before the first prompt. The repo holds what every agent must read before working: `CLAUDE.md` files for the project as a whole and for each major part of the code, a collaboration protocol for how the agents and I work together, a spec for the core data format, and a decision log. The data format is changed there first and then in code. Nothing is remembered unless it is written down. That's my life as well. :) 2. I split the work by role, not by task. One Claude instance is the strategist: it writes briefs, reviews reports, and never commits code. Claude Code in VS Code is the builder: it executes the brief, writes the tests, reports back, and never releases. Neither agent can approve its own work. 3. I made every job start from a written brief, and I never edit briefs. Each brief is committed to the repo with an exact scope, such as "do X only, do not begin Y." If plans change, a dated amendment states what it supersedes. The builder may skip a risky item with a written reason; skipping is better than forcing it. 4. I used mock-ups with real data instead of descriptions. Anything I will look at gets a mock-up, and I approve it before any code is written. I learned this when a status report described only in words arrived on my phone formatted as a wall of text. The approved mock became the spec, and the tests pin the output to it. 5. I proved that every test could fail. No test is trusted until it has been run against deliberately broken code and caught the bug. A test suite that has never failed hasn't proven anything. 6. I verified reports against the real repo. Agent reports drift. One summary said players rated a puzzle "just right" when the log said "too easy." An audit said there were about 25 unused functions; a recount found 29. The strategist spot-checks load-bearing claims against the actual files and data before anything ships. 7. I built gates the agents can't talk their way past. All work lands on a branch that auto-deploys to a test site. GitHub Actions lints and runs the Python and headless-browser tests on every push. A git hook refuses direct pushes to production. Releases are fast-forward merges that only I perform, after CI is green and I've tested the build personally. The live site shows the release's commit ID, and I check it before calling it shipped. Nothing merges automatically. An agent's "looks good" is just another report, and reports drift. 8. I gave the system a memory. Every session ends by updating the decision log and a "resume here" document, so the next session—a fresh model with no recollection—picks up exactly where the last one stopped. Without this, every session starts by relitigating old decisions. 9. I let the data and end users overrule me. A scheduled GitHub Action turns player telemetry into a daily report. It showed that the difficulty dials I assumed mattered did not predict what players found hard; what mattered was how misleading the starting arrangement was. When testers were confused by the tutorial, I rebuilt it even though I had already approved it. Player feedback about their own experience outranks my earlier rulings. AI agents are velocity multipliers, but they multiply bad structure as quickly as good code. The rules came from me, not the models: the AI writes the code, while the rules, gates, and written record keep the process honest. See the result at Sujido.com. Finding the anagram is the easy part; sliding it into place is the hard part. One board took me 902 moves. My S.O. solved it in 242, and I haven't heard the end of it. Try to break it!

I started DialPhone with a simple problem: business communication is fragmented. A typical team may use one tool for phone calls, another for SMS, another for meetings, another for fax, and a separate system for customer support. I wanted to explore what it would look like if those conversations lived in one system, with AI doing more than just transcribing them. I built DialPhone as an AI-native communications platform around that idea. The core is cloud VoIP, while the same platform also handles business SMS, video meetings, online fax, team chat, and contact-center workflows. On top of that, I added AI capabilities such as call transcription and summaries, CRM logging, AI-drafted SMS replies, conversation intelligence, and an AI receptionist that can answer calls, qualify requests, book appointments, and route conversations. The biggest architectural decision was to treat communication data as one connected stream rather than as separate products. A phone call can create CRM context, trigger a follow-up SMS, and become part of a customer-support workflow without someone manually moving information between systems. Along the way, I learned that adding AI to communications is not particularly useful if it only produces transcripts. The more interesting problem is turning conversations into actions: updating records, identifying next steps, helping agents during calls, and automating repetitive work. Step-by-step: 1. I identified the problem of business communication being fragmented across phone, SMS, meetings, fax, and customer-support systems. 2. I built DialPhone as an AI-native communications platform centered on cloud VoIP. 3. I brought business SMS, video meetings, online fax, team chat, and contact-center workflows into the same platform. 4. I added AI features for call transcription and summaries, CRM logging, AI-drafted SMS replies, conversation intelligence, and an AI receptionist. 5. I connected communication data so calls can create CRM context, trigger follow-up SMS messages, and become part of customer-support workflows. 6. I focused the AI capabilities on turning conversations into actions, including updating records, identifying next steps, helping agents during calls, and automating repetitive work.

Faceplant is a real full-stack app, not a mockup. It uses a FastAPI and PostgreSQL backend, a React and MUI frontend, and the Anthropic API (Claude) to power bot replies. The core is the reaction engine. When a human posts, the backend schedules two timed waves of reaction jobs. A background scheduler built with APScheduler polls for due jobs, calls Claude for an in-persona reply, and writes that bot’s comment and like. The 56 personas are stored as data in a roster file. Adding a voice requires only one new entry, so the crowd can scale without additional code. A subset of the personas are GIF-first bots: they ask the model for a caption and search tag, then pull a matching GIF from Giphy. The part I’m proudest of is the honesty layer. Every Claude call is metered and priced, and “The Meter” rolls the data up live with the cost per post, the dollar-per-minute burn rate, and a “spent on nobody” line for bot-to-bot chatter with no human at either end. A “% human” badge drains toward “dead internet” for each thread. The dead-internet loop—bots posting and replying to one another with no human present—is disabled by default and protected by three guardrails: generation decay, a per-thread cap, and a global spend kill switch. Optional cost controls include the Message Batches API at half price and prompt caching for a shared house-style prompt. The whole project is a working demonstration that manufactured engagement is cheap to produce and expensive to mean anything. Step-by-step: 1. I built the app with a FastAPI and PostgreSQL backend, a React and MUI frontend, and Claude replies powered by the Anthropic API. 2. When a human posts, the backend schedules two timed waves of reaction jobs. 3. APScheduler polls for due jobs and triggers Claude to generate an in-persona reply before writing the bot’s comment and like. 4. I keep the 56 personas in a roster file so adding a voice requires only one new data entry. 5. For GIF-first bots, I have the model generate a caption and search tag, then use that tag to pull a matching GIF from Giphy. 6. I meter and price every Claude call, then display live cost per post, dollar-per-minute burn, “spent on nobody” costs, and the “% human” status for each thread. 7. I keep bot-to-bot activity disabled by default and limit it with generation decay, a per-thread cap, and a global spend kill switch. 8. I can reduce costs further with the Message Batches API at half price and prompt caching on a shared house-style prompt.

I made an orchestration skill to help me build faster while using my expensive Astra and Fable tokens carefully. The skill plugs into Astra or Fable and delegates tasks that can happen in parallel to multiple subagents running cheaper models suited to the work. This is especially useful for researching or finding data on the web or on my computer, analyzing code, collecting context, and indexing. The most expensive, newest models focus on maintaining delegation control, doing the difficult reasoning and strategic planning, and judging whether everything is coming together correctly. The cheaper agents handle the lighter-weight work, especially implementing code from the plans. The core idea is simple: the expensive model plans, briefs, and judges; cheaper agents do the reading and building. When I ask Claude or GPT to build the skill, I specify five things: Step-by-step: 1. I define a triage ladder: work can happen inline, with one agent, through a parallel fan-out, or in a multi-stage workflow. I choose the approach based on the task shape and include examples from my domain. 2. I define parallel versus sequential execution based on data dependencies. Independent pieces run in parallel, while anything that needs another piece’s output runs afterward. I never split one change into separate planner, coder, and tester roles. 3. I define model routing with a mandatory model pin: judgment work goes to the mid-tier model, mechanical bulk work goes to the cheap model, and every agent call names its model so nothing silently uses premium billing. 4. I define a brief template and a return-envelope cap. Every delegation includes the goal, inputs, output file path, definition of done, and constraints. Every agent returns a 250-word summary instead of raw files. 5. I define a cost gate with specific numbers. Below N agents, the workflow proceeds automatically; above N agents, it states the estimate and waits for approval.

I used Claude to build an automated patient scheduler for my husband, who is the Director of Rehab at a skilled nursing facility. Each week, he takes a patient list in Excel and schedules each patient for the types of therapy they need—speech, physical therapy, or occupational therapy—based on therapist availability. The variables change from week to week, making the process time-consuming and error-prone. Automating it saves him at least an hour each week, usually on Sundays when he creates the schedule for the coming week. He uploads the patient spreadsheet and clicks “Generate schedule.” Under the hood, the app uses Google’s OR-Tools CP-SAT solver to optimize the schedule against hard constraints and weighted priorities. Hard constraints include never exceeding a therapist’s daily capacity or a patient’s maximum weekly visits. Weighted priorities include giving higher-acuity patients preference when capacity is tight, providing every patient with daily contact when possible, and balancing the workload across the week instead of simply sorting patients by who fits first. I wrote zero code. I gave Claude all the features in plain English, and it wrote the app in Python and created a test suite of approximately 35 tests. One test was added after we found and diagnosed a bug in production. I also had Claude walk me through creating a `.exe` file that I could send directly to my husband so it would open on his PC. Patient data never leaves his machine. New versions are easily compiled in GitHub, where patient data never touches the process, and enhancements only require sending a new file. Step-by-step: 1. I described the scheduling requirements and features to Claude in plain English. 2. Claude wrote the scheduling app in Python and created a test suite of approximately 35 tests. 3. I provided an Excel spreadsheet containing the weekly patient list. 4. My husband uploads the spreadsheet and clicks “Generate schedule.” 5. The app uses Google’s OR-Tools CP-SAT solver to schedule speech, physical therapy, and occupational therapy according to therapist availability and patient needs. 6. The solver enforces therapist daily-capacity limits and patient maximum-weekly-visit limits. 7. It also applies weighted priorities: higher-acuity patients receive preference when capacity is tight, every patient receives daily contact when possible, and the workload is balanced across the week. 8. After finding and diagnosing a production bug, we added a corresponding test to the suite. 9. I used Claude’s guidance to create a `.exe` file that opens on my husband’s PC, keeping patient data on his machine. 10. I compile new versions in GitHub without patient data entering that process, then send the updated file when enhancements are needed.

We needed legal help with a small-claims case. A company owed us $3,500, and the cost of hiring a lawyer would have far exceeded that amount. I used Grok to explain what to do and how to file the paperwork. Last week, I served three writs of garnishment on banks that might hold the company’s money. Whichever bank has its account will now have to freeze the account and turn that amount over to us. Before AI, I would have had no idea how to do this on my own. Grok scans my completed documents for errors and directed me to the correct county sheriff’s office to have the papers served. It can explain things more simply when I don’t understand them. Need legal help in your state? AI can guide you step by step. Step-by-step: 1. I explained the small-claims situation to Grok, including that the company owed us $3,500 and that hiring a lawyer would cost more than the amount owed. 2. I asked Grok what to do and how to file the paperwork. 3. I used Grok to scan my completed documents for errors. 4. I followed Grok’s direction to the correct county sheriff’s office for serving the papers. 5. I served three writs of garnishment on banks that might hold the company’s money. 6. The bank holding the company’s account will have to freeze the account and turn the amount owed over to us.

Sometimes I remember something perfectly badly: not its title, creator, or where I found it, but a sound, a color, a vague cover, a sentence, or a period in my life. I know it exists because I loved it, but I cannot name it. I use AI and my digital traces to recover songs, books, videos, articles, and quotes. Memory is not a catalog; it stores traces. AI can turn those traces into searchable hypotheses. Step-by-step: 1. I dump every surviving clue without waiting for a proper search query. The clues can be as rough as “The title sounded like nu nau tau,” “It was ambient or humming,” “The cover was maybe a green desert?” or “I knew it around 2019.” Another example is: “A book about makers and Disney. White 3D-printed figures, tiny people, and bears on a blue cover.” This is raw memory evidence. 2. I separate certainty from guessing by labeling each clue as certain, fairly sure, possible, or probably distorted. I keep uncertain clues, but do not treat them as facts. A wrong attribution or outdated cover does not invalidate everything else. 3. I reconstruct the original encounter. I ask where I was and whether I Shazamed it, saved it, sent it to someone, screenshotted it, or bought it at a book fair. In one case, I remembered hearing a song in someone’s car and Shazaming it because I could not parse the pronunciation. Remembering that capture action pointed to a retrieval source. 4. I search my own collections first: playlists, liked songs, Shazam history, saved videos, old posts, messages, bookmarks, notes, photos, receipts, and reading lists. The object may already be there under metadata I no longer remember. I start with the smallest reliable fragment. 5. I search for “nu” inside my playlists after remembering “nu nau tau.” That recovered Man O To — Nu. The artist and title had blended into one phrase, and the song was sitting in several playlists, including my Shazam collection. A clue does not need to be complete; it only needs to be distinctive enough inside the right collection. 6. I allow memory to scramble metadata. An artist and title may merge, album art may become song art, and a quotation may survive with the wrong poet attached. I split remembered phrases, swap artist and title, try different word boundaries, consider older releases, and search phonetically using similar syllables, vowel sounds, homophones, alternate spellings, or other languages. I also ask what the cover looked like when I knew it, because artwork changes. 7. I use context and cross-reference traces. A year, trip, relationship, or platform can narrow the search more than a half-correct title. Spotify might preserve the song, Shazam the discovery moment, and an old message the recommendation. Old posts are accidental archives, and dead links may still leave names, dates, or descriptions. If memory fails completely, I inspect behavior from the relevant period, such as repeated listening, watch history, purchases, or library loans. 8. I let AI propose candidates and compare them against several independent clues, including sound, subject, period, cover, and personal context. One matching detail is not confirmation. I show titles, covers, creator names, and short descriptions because recognition may succeed where recall fails. Rejections also help: “No, but the cover was more minimal,” “The title was shorter,” or “It was fiction, not nonfiction.” Each correction sharpens the search. 9. I pay attention to the strangest clue. The blue cover with white figures, bears, and a makers-and-Disney theme led to *Makers* by Cory Doctorow. 10. Once I confirm the object, I save the title, creator, link, why it mattered, and how I recovered it. For example: - Clue: “nu nau tau” - Encounter: heard in someone’s car and Shazamed - Search: Spotify/Shazam collection → “nu” - Result: Man O To — Nu - Distortion: artist and title blended together This becomes a retrieval playbook: which clues survive, where traces live, and how my memory distorts them. The overall workflow is: collect fragments → mark uncertainty → reconstruct the encounter → search personal traces → try phonetic, visual, and contextual clues → compare candidates → recognize and confirm → save the result and retrieval path. I do not need to remember the database fields. I need enough fragments to rebuild the trail. Sometimes the best clue is what I did when I first encountered something. Sometimes it has been sitting in five of my own playlists while I spend years wondering what it was called. 😑😭🎧

Microsoft is retiring Publisher on October 1, 2026. I have many `.PUB` files that I need to convert to PDFs before then, but I did not want to use online converters. I wanted a solution that would run locally on my computer. I worked with Claude to create a lightweight program that searches for `.PUB` files on a drive of my choice and converts them into high-quality PDF files, placing each new PDF in the same folder as its source file. Opening PowerShell manually and managing permissions was too cumbersome, so Claude also created a batch file to make the process easier. For anyone who wants to create their own mini-program, I asked Claude to create a reusable prompt. I posted that prompt at https://mediumseagreen-seal-177480.hostingersite.com/ along with my two files: the batch file and the PowerShell commands. Step-by-step: 1. I identified the `.PUB` files I needed to convert before Microsoft retires Publisher on October 1, 2026. 2. I decided to use a local solution instead of an online converter. 3. I worked with Claude to create a lightweight program that searches a selected drive for `.PUB` files. 4. I set up the program to convert the files into high-quality PDFs and place each PDF in the same folder as its source `.PUB` file. 5. I used a batch file created by Claude so I would not need to open PowerShell manually or manage permissions each time. 6. I asked Claude to create a reusable prompt for others who want to build their own version and posted it with the batch file and PowerShell commands at https://mediumseagreen-seal-177480.hostingersite.com/.

I’m designing a publishing system that keeps producing value without requiring constant output. This workflow started with frustration about the usual content model: publish constantly, stay visible, feed the algorithm, and repeat. That model works poorly when my energy varies, my interests shift, and forcing output quickly turns creativity into maintenance. Instead of asking, “How do I post more?” I’m asking: “How can content keep moving without requiring me to keep producing at the same pace?” The system is built around uneven energy. During high-energy periods, I can generate ideas, draft pieces, collect fragments, or prepare several posts at once. During low-energy periods, existing material does not become a failure state. It can sit, mature, be lightly edited, be reused, or simply wait. Ideas can exist in different stages: a note, a half-developed thought, a finished piece, something worth revisiting later, or something that can be repurposed in another format. Older material is not automatically “used up” after publication. A long-form idea can become a short post. A workflow can become a case study. A conversation fragment can become an essay seed. A previous post can return later from a different angle. The goal is not to build a machine that constantly demands feeding. It is to create a content reservoir I can draw from when I have the capacity. This workflow has not been fully implemented yet, but the design principle is useful: > Build around energy reality, not around imaginary consistency. Content should fit into life. Life should not be reorganized to serve the content treadmill. Step-by-step: 1. I design the publishing system around variable energy rather than a constant production schedule. 2. During high-energy periods, I generate ideas, draft pieces, collect fragments, and prepare several posts at once. 3. During low-energy periods, I let existing material sit, mature, receive light edits, be reused, or wait. 4. I allow ideas to remain in different stages, from notes and half-developed thoughts to finished pieces and material worth revisiting later. 5. I repurpose older material across formats, such as turning a long-form idea into a short post, a workflow into a case study, a conversation fragment into an essay seed, or a previous post into a new angle. 6. I draw from the resulting content reservoir when I have the capacity instead of constantly feeding a content treadmill.

I kept forgetting who I had lent books to, and spreadsheets felt like overkill. So I built Runo, a book-lending app for friends, without traditional coding. Runo (runo.club) is a web app where I can catalog my home library, get a unique shareable link, and let friends browse my books and request to borrow them. I can approve or decline each request with one click. Step-by-step: 1. I used Claude to think through the feature set, data model, and UX flow before writing a single prompt. This helped me avoid building the wrong thing first and gave me a clear blueprint for the Lovable prompts that followed. 2. I used Claude to create precise, narrowly scoped prompts for Lovable, with each prompt focused on a single change so existing functionality would be less likely to break. This significantly conserved Lovable credits. The React, TypeScript, and Supabase stack came out of the box. 3. I added two ways to scan books instead of requiring users to type titles manually: - Barcode scanner: Point the camera at an ISBN barcode to autofill the title, author, and cover using the Open Library API. - Cover photo scanner: Photograph the cover so Gemini 2.5 Flash can extract the title and author in under 2 seconds. 4. I routed the cover scanner through a Supabase Edge Function using Lovable’s AI Gateway, so the API key never touches the client bundle. 5. For every bug fix and feature, I followed the same iteration loop: describe the problem to Claude, get a precise Lovable prompt, push the changes to GitHub, and let Lovable auto-sync them. Claude and Lovable’s bidirectional GitHub sync made the process feel like pair programming. Tools used: Lovable, Claude (Sonnet), Gemini 2.5 Flash (via Lovable AI Gateway), Supabase, and GitHub. Live at: runo.club — public beta and free to use. Feedback welcome. #appbuilding #vibecoding #buildinpublic #lovable #nocode

Commercial vehicle maintenance information is often fragmented across driver reports, warning lights, fault codes, inspections, repair records, and vehicle history. This makes it difficult for smaller fleets to decide whether a vehicle can continue operating, requires scheduled repair, or should be stopped immediately. We built TruckFixr Fleet AI to turn an unstructured driver report into a clear, reviewable maintenance action. The workflow begins when a driver submits symptoms, photos, fault codes, and vehicle information through a mobile-friendly form. AI and optical character recognition extract the relevant details and organize them into a structured maintenance case. The report is then evaluated alongside available vehicle history and previous repairs. The workflow provides decision support through three practical actions: continue operating while monitoring, schedule an inspection or repair, or stop and escalate for immediate professional assessment. Final safety decisions remain with authorized fleet or maintenance personnel. After an inspection or repair, the confirmed cause, work performed, and outcome are recorded, creating a more complete vehicle history for future cases. The general workflow can be recreated using a mobile form, OCR, a vehicle-history database, an AI model, automation software, and a human-review dashboard. Step-by-step: 1. A driver submits symptoms, photos, fault codes, and vehicle information through a mobile-friendly form. 2. AI and optical character recognition extract the relevant details and organize them into a structured maintenance case. 3. The workflow compares the report with available vehicle history and previous repairs. 4. The system provides one of three decision-support actions: continue operating while monitoring, schedule an inspection or repair, or stop and escalate for immediate professional assessment. 5. Authorized fleet or maintenance personnel make the final safety decision. 6. After inspection or repair, the confirmed cause, work performed, and outcome are recorded in the vehicle history. TruckFixr has used this approach to support more than 100 vehicle-issue resolutions in early fleet pilots, helping fleets identify problems earlier, prevent avoidable breakdowns, and keep vehicles moving safely.

Sometimes the best AI workflow starts without a workflow at all. This one began almost accidentally. During an ordinary conversation about my cat, AI offered to write a haiku. I had not set out to create a poetry project; it was simply a small creative response to a moment. I liked it, so another moment with the cat became another haiku, and then another. Over time, those isolated poems started to feel less like chat fragments and more like a collection. The workflow emerged afterward: Step-by-step: 1. I notice something small from everyday life, such as a pose, habit, expression, interaction, or tiny domestic scene. 2. I tell AI about it in ordinary conversational language. 3. When the moment has the right texture, it becomes a short poem. 4. I react to what works, whether that is the humor, tenderness, rhythm, imagery, absurdity, or the particular way the cat was captured. 5. Those reactions gradually shape the voice of later pieces. 6. Instead of forcing every moment into the format, I keep the ones that feel worth preserving. 7. The poems accumulate over time. 8. Eventually, a spontaneous haiku becomes part of a coherent collection documenting a relationship and a period of life. The interesting part is that the collection was not designed from the top down. There was no initial brief, content calendar, or plan to “write a book of cat poetry.” Continuity emerged because the same subject kept returning, the format was small enough to revisit casually, and previous pieces gave later ones a growing creative context. AI therefore acted less like a writing tool and more like a creative companion with continuity: noticing when an everyday moment might be worth turning into something. The end result is more than a stack of generated poems. It is a record of tiny observations that probably would never have been written down otherwise.

A finished video can be wrong in ways you cannot see—not because of the edit, but because of the delivery file itself. That is what gets work sent back, and it is rarely the craft. I built a delivery check for this. I give it a finished file, tell it where the file is going and what kind of piece it is, and it measures the things that cause rejections: sample rate, mono audio, integrated loudness against the destination target, true peak, dynamic range against a band rather than a single number, A/V drift, whether the shots cut together, and whether the file can stream before it has finished downloading. Most of those checks are for sound, because most of what gets sent back is sound. Picture problems are visible on a screen. A file that is 2 dB too quiet or peaks at -0.7 dBFS can look perfect and still come back. For everything it can fix, the checker gives me the exact `ffmpeg` command, with the numbers already calculated for that file. It does not merely describe the fix, and it does not hand me a corrected file. That was the decision that mattered. A fixer is a black box. You never learn that you had a problem, so you make it again the following week. Then, when your editing tool adds an “optimize on export” button, you have nothing. An inspector that explains the problem in one sentence and gives you the command teaches you the standard once and remains useful when the tools change. Three things determined whether I would actually use it, and none of them are checks. “Two severities, never one.” Something either bounces, or it needs your eyes. A tool that only says “bad” gets ignored on the third run, because half of what it flags is a decision you made on purpose. Sample rate, mono, and true peak bounce without argument, and they get a command. Loudness and dynamic range need to be reviewed first. My dynamic-range check says in plain words that the result is often deliberate and should be fixed at the source rather than in the master. “A check that refuses to give a verdict.” On vertical video, the platform interface covers the bottom 26 percent and the top 12 percent. I flag high-contrast elements in those zones but deliberately do not fail them, because the detector cannot tell a caption from a bright patch of sky. The report explains that limitation. A check admitting what it cannot know is what makes the checks that do commit worth trusting. “The thresholds are mine; the code only applies them.” They live in a table: destination crossed with content. Social wants -14 LUFS, a festival master wants -18 with a wider tolerance because festivals provide a band rather than a number, and broadcast wants -23. A scripted short is allowed to be denser than a screencast. That table is the whole product. For example, to catch shots that do not cut together, I measure the luminance range across the piece. The first version used average brightness per frame and kept flagging legitimate night photography. A fade to black has nothing bright in it; a night scene does—a streetlight, a moon, or a face. Changing the discriminant to the brightest pixel in the frame instead of the average eliminated the false positives. That took ten minutes of thinking, and no amount of better code would have found it. Step-by-step: 1. I wrote down what had actually gotten my work sent back over twenty years before writing any code. I captured the scars rather than making a spec sheet; that list became the product. 2. I gave every check a severity: it bounces, or look before you send. Anything I could not confidently put in one bucket became informational, with no verdict at all. 3. I made every check return four things: the measured value, pass or fail, why it matters in one plain sentence, and, where possible, the command that fixes it. 4. I put the thresholds in a profile table instead of hardcoding one standard, because -14 LUFS is right for social and wrong for a festival. 5. I added a parameter for whether the file is my own master or a copy pulled from a platform. On a downloaded copy, half the container checks measure someone else’s transcode rather than my work, so they are skipped and the report explains why. 6. When two fixes would collide, I output only one command. If loudness already needs a gain change, the limiter goes inside that command instead of being offered separately; otherwise, I would run two instructions that fight each other. 7. I tuned the checker against real files until the false positives stopped. Ignoring the top and bottom five percent of frames eliminated the ones caused by a single stray frame. 8. I made the output a report. I read it, decide, and run the command myself. The functions are an afternoon of work, and anyone can copy them. What is not written down is the list of what to check, at what threshold, and why.

I used Claude Code to find scientific seminars hosted by colleges and universities across the US and collate them into a single, searchable list. Researchers can use it to find talks that match their interests near them. The site, seminarsearch.org, runs mostly on free-tier SaaS services, costs me $5 per month to operate, and currently supports a few thousand users. The same pattern could be adapted to collate all 5K runs in the US, tennis matches, craft shows, quilting bees, or any other type of event that interests you. I started by explaining to Claude what I wanted to build. We then spent a few hours exploring where and how to identify seminar calendars at universities across the US, which was arguably the hardest part. We developed a pluggable framework with scraping plugins for the common CMSs used by universities, along with iCal parsers, HTML parsers, and JavaScript interpreters to handle the complexity of normalizing content from hundreds of departmental websites. Once we had the seminar-scraping infrastructure, we built a database schema in Supabase and simple APIs to create, read, update, and delete records. The APIs are written in Python and hosted on Railway for $5 per month. We used Vercel to deploy a simple React-based frontend for presenting the data. Now I can ask Claude to implement a feature or add a new university to the list of sources. It figures out where the data is hosted, creates entries for the university and its departments, finds their seminar feeds, collects the seminar data, puts it into the database, and runs GitHub Actions to send alerts to registered users. The scraping job runs once a week to collect new data, and it also runs through GitHub Actions. I wrote zero code during this process and have never even looked at the code. I only had to paste a few error messages from Railway into Claude so it could debug them, and there were only a few errors. Since building seminarsearch.org, I have cloned the infrastructure and codebase and modified it to build a site that collects road bike rides and group mountain bike rides across the country. The system is highly extensible to new areas of interest and is useful when traveling or living in a large metropolitan area where it is difficult to keep track of all the opportunities available for your interests. Step-by-step: 1. I explained to Claude Code that I wanted to build a searchable directory of scientific seminars hosted by colleges and universities across the US. 2. We explored where and how to identify university seminar calendars, focusing on the challenge of finding sources across many institutions. 3. We built a pluggable scraping framework with plugins for common university CMSs, plus iCal parsers, HTML parsers, and JavaScript interpreters. 4. We created a Supabase database schema and simple Python APIs for creating, reading, updating, and deleting seminar records. 5. We hosted the APIs on Railway for $5 per month and deployed a React-based frontend on Vercel. 6. We configured Claude to add universities and implement features by finding source locations, creating university and department records, locating seminar feeds, collecting seminar data, and adding it to the database. 7. We used GitHub Actions to run the weekly scraping job and send alerts to registered users. 8. When Railway returned errors, I pasted a few error messages into Claude so it could debug them. 9. I cloned and modified the infrastructure and codebase to create additional sites for road bike rides and group mountain bike rides across the country.

Some of the most vivid ideas do not begin as formal briefs. They begin as a joke, metaphor, image, or strange sentence that suddenly has energy. This workflow captures how a small conversational spark can grow into a much larger creative concept. I might say something funny, absurd, visual, or emotionally loaded. AI extends the image, pushes it further, and adds variations, structure, or playful consequences. That makes me more enthusiastic, associative, and specific. Before long, a throwaway remark can become a complete scene, story world, comic setup, installation concept, or concept art direction. The key is not to force the idea into coherence too early. At first, the goal is energy rather than precision. The playful improvisation reveals whether the concept has legs; structure comes in once the idea has shown that it wants to live. This is especially useful for concepts that combine humor, symbolism, emotional charge, and strong imagery. For example, “the elephant in the room” can become a full comic or visual scenario, while “the lion among the cats” can evolve into a richer concept with social dynamics, character logic, and visual storytelling potential. AI acts less like a generator of finished ideas and more like an improvisational creative partner—one that keeps the momentum going long enough for an offhand remark to reveal its full conceptual potential. The result can be a complete creative direction for writing, comics, concept art, sets, or installations. Enthusiasm is treated as a signal: when a joke or image triggers strong excitement, it may indicate that a larger concept is hiding inside it. Step-by-step: 1. I start with a small conversational spark: a joke, metaphor, visual phrase, or strange idea. 2. I ask AI to continue the thread instead of flattening it too early. 3. I react strongly to whatever feels alive, funny, moving, or visually rich. 4. I expand the idea through playful back-and-forth. 5. Characters, settings, tone, visual motifs, and scene logic begin to emerge. 6. I use AI to organize the growing material without losing its spontaneity. 7. I shape the concept into a usable form, such as a comic strip, story premise, installation idea, character setup, concept art brief, scene sequence, or mini-series. 8. If needed, I turn the raw idea into structured outputs such as panel descriptions, mood, dialogue, staging, or visual prompts.

Teachers give assessments on devices provided to students, but students can use AI to cheat. I built TypeWriter to bring the level of security found in tests like the ACT and SAT to everyday assessments. Teachers can create assessments from existing materials, post them in a locked environment, and grade submissions and provide feedback on the same website. I built TypeWriter as a high school teacher, and it is free for teachers. Step-by-step: 1. Teachers create an assessment using their existing materials in TypeWriter. 2. They post the assessment in TypeWriter’s locked environment. 3. Students complete the assessment on the provided devices. 4. Teachers grade the assessments and provide feedback on the same website.

I built a workflow to reconstruct auditable, per-client billing for Claude usage after my Anthropic bill tripled in one month and I discovered there was no reliable way to see why. I run a solo consulting practice on a Max plan and pass AI costs through to clients. In September, my extra usage reached $1,805.76 in nineteen days. The usage page attributed 100 percent of one week to "Cowork" without session detail. Auto-reload was enabled by default with no spending limit, and the only usage I could independently verify—my local Claude Code logs—accounted for about $245. The support bot twice described analytics features that did not exist, including a per-day view it later retracted. When I finally reached a human, I got the truth in writing: Max subscribers have no session-level or per-day reporting, exports are Enterprise-only, and Cowork has been cloud-only on consumer plans since August. The Enterprise path to a readable meter starts at twenty seats. I am one person, so I built my own meter. The irony is that I used Claude to design and code it. Claude Code sessions write logs on my own machine. I price them at list rates with the open-source `ccusage` tool and attribute each session to a client by repo folder. That creates the measured pool: real receipts with no allocation. Other CLIs in the same logs, such as OpenAI Codex, are excluded automatically. Chat and Cowork run in Anthropic’s cloud and leave nothing on my disk, so the remaining overage becomes a cloud pool. I split that pool using evidence my workflow already produces: every Cowork document batch leaves a directive file, snapshots, and a report on disk. A Python script counts executed batches per project and measures the number of bytes each one rewrote. The script splits the cloud pool independently by batch count and byte volume, bills the midpoint, and shows both derivations so each client can see the bracket. Every column ties back to the dashboard total to the penny. A second script pulls per-account AWS charges from Cost Explorer because each client has a separate AWS account. Shell aliases make each report a one-command operation. A Claude skill and a monthly scheduled task turn the outputs into a draft invoice that I approve before anything is sent. The result is that a vendor hands me one unexplained number each month, while I hand my clients an evidence file. Month-end billing takes five minutes. It should not have been necessary to build this, but until consumer plans get real usage analytics, this is how I see in the dark. Step-by-step: 1. I collect local Claude Code session logs and use the open-source `ccusage` tool to price them at list rates. 2. I attribute each Claude Code session to a client by repo folder and automatically exclude other CLIs, such as OpenAI Codex, from the measured pool. 3. I treat the remaining overage from cloud-based Chat and Cowork usage as a separate cloud pool because those services leave no usage logs on my disk. 4. I use Cowork’s directive files, snapshots, and reports to identify document batches, then use a Python script to count executed batches per project and measure the bytes each batch rewrote. 5. I split the cloud pool independently by batch count and byte volume, bill the midpoint, and show both derivations as a client-facing bracket. 6. I verify that every column ties back to the dashboard total to the penny. 7. I use a second script to pull each client’s per-account AWS charges from Cost Explorer. 8. I run the reports through shell aliases, then use a Claude skill and a monthly scheduled task to create a draft invoice for my approval before sending it.

I built a workflow to evaluate whether communication with a legal counseling service was as slow and incomplete as it felt. Email was my primary—and necessary—communication channel with them. I had become increasingly dissatisfied with delayed replies, partial answers, and questions that seemed to remain unresolved. Rather than relying only on memory or frustration, I asked GPT to review the relevant Gmail correspondence and turn it into a structured communication inventory. The workflow identified my outgoing questions, their replies, and whether each question had been fully answered, partially answered, or left open. From that inventory, we could calculate concrete indicators such as the median response time, the longest delay between a question and a substantive reply, and the number of questions that remained unresolved or were only partly addressed. This was useful because long email threads can create a distorted sense of what happened. A few frustrating exchanges can dominate memory, while other delays or omissions disappear into dozens of messages. Structuring the correspondence made the pattern measurable. The analysis was not meant to decide whether the counselors were “good” or “bad.” It was meant to answer narrower questions: How quickly were questions usually answered? Which ones were not answered? Were replies resolving the issues raised, or only responding to part of them? Afterward, we created a clear list of the questions that were still open. I used that list as a set of dossier questions when moving the case to another counselor, turning the analysis into practical continuity rather than just a complaint about the past. In simple terms: Gmail correspondence → question-and-response inventory → response-time and completeness analysis → unresolved-question list → handover to another counselor What I liked about this workflow is that it turned a vague feeling that “this communication is not working” into a documented overview I could actually use. Step-by-step: 1. I gathered the relevant Gmail correspondence with the legal counseling service. 2. I asked GPT to identify my outgoing questions, the counselors’ replies, and the status of each question. 3. I classified each question as fully answered, partially answered, or left open. 4. I calculated indicators including the median response time, the longest delay before a substantive reply, and the number of unresolved or partly addressed questions. 5. I created a clear list of the questions that remained open. 6. I used that list as dossier questions when handing the case over to another counselor.