My family has a group text. For about two years its main job was four people sending each other the same true crime documentary, a day apart, each one sure they found it first. Someone would ask if it was on Netflix or HBO Max. Someone would say Hulu. Someone would say they were already halfway through it. Nobody wrote any of it down.

So I built a place to write it down. It is called wen streaming? and it lives at wenstreaming.com.

What it is

A watch club. One shared list for a group of people, which the site calls a crew. Everyone in the crew can add a movie or a show, tap in for a night, rate what they watched and argue about it in the comments. The list belongs to the crew, and so does the history. You have a profile, but the crew is the account that matters.

Every Thursday morning the site sweeps the big services, Netflix, HBO Max, Hulu, Prime Video, Disney+, Apple TV+, Peacock, Paramount+ and PBS, and files what is new and what is coming. Then one email goes to the crew: what landed, what people added, what anyone said about it. One email a week. No notifications, no daily nag. Thursday because releases cluster from Wednesday to Friday and most of the watching happens on the weekend, so the list is ready before anyone has to decide.

There is a cast. Sixteen objects, and only objects: a scarlet television with a bent antenna, a clapperboard, a question mark that carries its own dot around like a pet. Each one has exactly one opinion about movies. You pick one when you join. None of them could be mistaken for a person, which was the rule from the start.

The how-it-works page is drawn as a treasure map. Six stops, from your crew at the start to the scoreboard at the end of the week: most watched, harshest rater, biggest adder.

Why

True crime nights. My sister, my mom, my aunt and my cousin. Five people who all watch the same kind of thing and had no shared place to keep track of it.

The first version was a Notion database I kept by hand. It worked, which is how I knew the idea was fine and the tool was wrong. My family was never going to open Notion.

How it is built

The plain-language version.

The app is built in Lovable, a tool that writes the code from written instructions and hosts the result. Under it is Supabase, a hosted Postgres database, which holds the titles, the crews, the ratings and the comments. Posters, summaries, trailers and which service carries what come from TMDB, the community-run movie database. Scores from IMDb, Rotten Tomatoes and Metacritic come through OMDb, a second database that collects those. A scheduled job runs the Thursday sweep against TMDB's release data and refreshes where each title is streaming.

Then there is a second pass the feed cannot do. Once a week Claude reads the service press pages, the roundups and the true crime and documentary threads, and catches what the database misses: the small documentary that only got a press release, the British series on a channel nobody here has heard of, the library title that quietly came back. Those get filed to the same list, and the pass ends with a short list of titles I probably missed. The feed finds the obvious, and a well-briefed reader finds the good stuff.

The Notion database, 334 titles, was imported on day one, so the list never had an empty state to apologize for.

What is next

Crews for anyone. My family is the first crew. Anyone can make an account today, and the next job is making it dead simple to start your own crew and invite people with one link.

Where-to-watch links. Each title page will list the services that carry it in plain text and link out to them, so the list answers the only question anyone has at nine on a Friday night.

Crew plus, later. There is a shape for a paid tier and nothing behind it yet, so there is no pricing and no button. The free version is the product. A paid one shows up when there is something worth paying for.

If your group text does this too, start a crew at wenstreaming.com. Bring your people and a bad opinion.

BATNA stands for Best Alternative to a Negotiated Agreement. In plain words, it is the best thing you can do if this deal falls through. Your plan B, with a name that sounds like a minor Batman villain.

It is easier to see than to define, so here is a made-up example from the concert business. None of these people, rooms or numbers are real.

The setup

Picture a 1,200 cap room in a mid-sized city. "Cap" is short for capacity, so the room holds 1,200 people. The talent buyer wants a band for a Saturday in October. A talent buyer is the person at a venue or festival who hires the acts.

On the other side of the call is the booking agent. A booking agent is the person who sells an artist's live shows and negotiates the fee. The agent works for the artist.

The buyer offers $12,000 as a guarantee. A guarantee is a flat payment the artist gets no matter how many tickets sell. The agent asks for $18,000.

Find each side's plan B

The agent's BATNA: a 900 cap room across town has offered $13,000 for the same night. If this deal dies, the band still has a $13,000 Saturday.

The buyer's BATNA: a regional act that would play that Saturday for $6,000 and probably sell around 700 tickets. A decent night.

What the BATNA tells you

The agent should turn down anything under about $13,000, because walking away pays better. The buyer can do similar math. If the bigger band sells 1,100 tickets and the regional act sells 700, the gap in ticket and bar money shows how high the buyer can go before plan B starts to look better.

So the deal, if there is one, lives between $13,000 and the buyer's ceiling. Say they land at $15,000 plus a bonus if the show sells out. Both sides beat their plan B. Everyone goes home and eats a sandwich.

Three things to take with you

1. Know your own BATNA before you pick up the phone. Write it down.
2. Make a fair guess at theirs. An agent who knows the buyer's October calendar is empty is holding useful information.
3. Improve your BATNA and you improve your deal. The agent who takes one more call to firm up that second offer has done more for the fee than any clever line at the table.

A BATNA is a floor, and you still want to aim higher than the floor. It also lets you be friendly. When you know you will be fine either way, there is no need to bluff. You can be a decent person to deal with, and in a small industry that reputation walks into the next room before you do.

Next time you negotiate anything, a salary, a used car, whose turn it is to do the dishes, ask one question first. If this falls apart, what do I do instead?

For six weeks I treated my terminal like a kitchen I kept remodeling. New skill on Monday, a fresh MCP server by Wednesday, a config file that had grown opinions of its own by Friday. Then I stopped, and the thing has now gone a full month without me touching it. That is the only benchmark I trust: not how clever a setup looks on day one, but whether you leave it alone once the novelty wears off.

Here is what survived.

One rule did the filtering

Every piece of the setup has to save me something I can name out loud. Not "it feels faster." Something like "I no longer retype my project conventions three times a session" or "I stopped hunting for where I saved that output."

That rule killed about half of what I had installed, and it killed the half I was most proud of. The clever stuff tends to be the stuff you built for the fun of building it. Worth admitting, then worth deleting.

The instructions file carries the load

The single highest-leverage piece is a global instructions file. Mine covers four things and nothing else:

• Who I am and how to talk to me. Direct answer first, no preamble, close with next steps.
• Where things live. One output folder, named explicitly, so nothing lands on my Desktop.
• What the machine is. Model, OS, drives, the connectors I use. This kills an entire category of confidently wrong advice.
• Safety rules. No deletes without approval in that same message. Show a dry run before any bulk move.

Earlier versions of this file ran twice as long and worked half as well. Long instruction files get skimmed the same way long emails do. Cut to the rules you would enforce on a bad day and the rest can live in skills.

Skills: fewer, sharper, and mostly about my own preferences

Skills earn their keep when they encode something a general model cannot guess. Mine fall into two buckets.

The first bucket is voice and standards. Writing rules, a lock on how my career credits get described, a tone default for chat. These prevent the slow drift where every draft comes back sounding like a press release from a company that does not exist.

The second bucket is workflow. A skill that forces a live documentation call before answering anything about how an app behaves. A skill that checks my real macOS settings before prescribing a click path. Both exist because I got burned by confident answers about menus that had been renamed two versions ago.

What I deleted: anything that was a nicely formatted checklist I could have kept in a note. If a skill does not change behavior, it is documentation wearing a costume.

MCP servers: three stayed, several went

The keepers do work that the model genuinely cannot do from inside a chat window.

• A documentation fetcher. Pulls current library and API docs on demand. This one has the highest hit rate of anything I run, because training data ages and APIs do not care.
• A real shell with persistence. Long-running processes, REPLs, and file work that outlives a single turn.
• My task system. Read and write access to the notes app where my actual to-do list lives, so nothing has to be copied by hand.

The ones I removed were all variations on the same mistake: a server that sounded useful, connected fine, and then sat unused for a month while its tool definitions ate context on every single turn. That last part is the hidden cost nobody mentions. An MCP server you never call is not free. It is a tax on every request.

Permissions, done once, properly

I spent an afternoon watching which commands prompted me over and over, then allowlisted the read-only ones. Listing files, checking git status, reading system version, searching with ripgrep. Writes and deletes still prompt, every time, no exceptions.

The result is a session where the interruptions mean something. When a prompt appears now, I read it, because it is no longer the fourth one in two minutes asking permission to run ls.

The month of quiet

The tell that a setup is finished is boredom. Nothing surprising happens. Output lands where I expect, the voice comes back sounding like me, and the safety rails catch the one stupid thing I try per week.

If your setup still feels exciting, it is probably still a project. Give it another month and see what you stop opening.

Running something you think I should test? Send it over. I will break it in public.

When I tell people I work in AI quality control, I usually get a polite nod and a quick change of subject. Fair. It sounds like a job title from a sci-fi movie. The day to day is much more familiar. Picture an editor, a teacher and a video game playtester sharing one desk.

Here is the short version. An AI model writes answers. Somebody has to check whether those answers are any good. That somebody is me.

Step one: read the answer. All of it.

A person asks a question, the model responds, and I read the response the way a careful editor would. Slowly. Twice if needed. Skimming is how bad answers sneak through.

Step two: grade it against a rubric

A rubric is a written scoring guide. It lists what a good answer needs to do and how much each part counts. Teachers use rubrics to grade essays. Evaluators use them so two different people can look at the same answer and land on the same score. Without one, grading turns into vibes, and vibes are hard to repeat on a Tuesday.

Step three: check what a score can miss

I look at four things on every answer.

• Tone. Does it sound like a helpful person, or like a terms of service page?
• Clarity. Could a busy reader follow it on the first pass?
• Usefulness. Did it answer the question that was asked, or a nearby question that was easier?
• Trust. Are the facts right? Does it say so when it does not know? A confident wrong answer is the most dangerous kind, because it looks exactly like a right one.

Step four: write test prompts

A prompt is the question or instruction you give the model. Part of the job is writing prompts built to find weak spots. Tricky wording. Missing information. Two instructions that pull in different directions. It is a lot like playtesting a game by running into every wall to see which ones you can clip through.

Step five: explain why a good answer is good

This is the part people do not expect. Choosing the better of two answers is only half of it. I also have to write down the reason in plain language, clearly enough that someone else could apply the same reasoning tomorrow. A note like "it felt better" helps nobody. A useful note reads more like "it answered the question in the first sentence, gave the steps in order, then flagged the one risk."

That is the day. Read, grade, poke, explain, repeat. It is patient work, and I love it. Every clear note helps the next answer land a little better for someone with a question and not much time.

Curious about this kind of work? The core skills are reading closely, writing clearly and being fair. No wizard robe required, though I would wear one.

I have put an absurd number of hours into games where the last thirty minutes involve a very large person on a throne explaining why order requires obedience. The speech is always the same. The throne is always too big. And the health bar, every single time, is longer than you expected.

Authoritarianism runs that script with a straight face. Once you notice the pattern, the whole thing gets less frightening and a lot more workable, because boss fights have rules and rules can be learned.

Phase one is always the speech

The boss monologues because the monologue is the weapon. It wants you certain that the outcome is decided, that resistance is a formality, that the smart people already made their peace. Games telegraph this on purpose. You are supposed to feel small in the cutscene so the fight lands harder.

The counter is boring and effective: keep moving during the speech. Nothing said on a throne changes what a person can do with a Tuesday afternoon, a phone, and a list of neighbors.

Nobody solos the boss

Here is the mechanic that every party learns the hard way. You cannot out-damage an end-game encounter alone. The math does not work. You can be the best-geared player on the server and you will still get flattened in phase two while everyone watches from the entrance.

Organizing is a party mechanic. The person who knows every city council member is your tank. The person who reads the budget documents is your mage. The person who feeds forty people out of one kitchen is your healer, and healers win more fights than damage dealers ever admit.

Solo heroics feel good and accomplish very little. Five people who show up every month beat one person who burns out beautifully.

Clearing adds is the actual job

The boss does not fight you directly most of the time. It summons adds. Small, tedious, endless: a rule change nobody reads, a board seat nobody contests, a library funding vote at 6pm on a Thursday in February.

Adds are unglamorous and clearing them is how the fight gets won. A school board meeting has no soundtrack. It has bad coffee and a sign-in sheet, and the person who shows up gets a vote on what happens to ten thousand kids. That trade is absurdly in your favor and almost nobody takes it.

The boss counts on you finding the small stuff too boring to bother with. That is the strategy. That is the whole strategy.

Gear check

Your loadout for this one is short.

• A local information source you trust. Preferably one with a reporter who attends meetings in person.
• Three phone numbers. Your representatives, saved, so calling takes eleven seconds instead of a research project.
• A recurring donation you barely notice. Small and monthly beats large and once, because organizations budget on reliability.
• One skill you already have. You can drive, cook, design, code, translate, or hold a clipboard. Every one of those is in demand somewhere within five miles of you.

That is it. No purity test, no reading list, no hoodie.

Peace and love is a build, not a vibe

There is a version of this work that runs entirely on anger, and it burns out its players by the second expansion. Rage is a great opener and a terrible sustain stat.

The builds that last run on something warmer: liking your neighbors, wanting the block to be nice, enjoying the potluck. Joy is not a distraction from the fight. It is the resource regeneration that lets you show up again next month, and next month is when most of it gets decided.

Also worth saying plainly: the people on the other side of this are not a final boss. They are mostly scared, mostly sold a story, mostly reachable. The boss is the machine that profits from keeping them scared. Aim at the machine. Be decent to the people.

The part where you win

Boss fights end. Not cleanly, and usually not in one session, but they end. Somebody figures out the pattern, posts the strategy, and suddenly a fight that wiped a hundred groups becomes something a coordinated party clears on a weeknight.

We are in the part where the strategy is getting written. Show up, clear adds, keep the healer fed, and try to enjoy the run. It is a long one.

Got a local group worth shouting out? Tell me about it.

Hey now. Welcome to piddonkadonk. I built a website. It has a blog, a donkey-adjacent name, and more scarlet than a doctor would recommend.

Here is what this place is. Three things live here.

• A video blog. Me, a camera, and whatever I am excited about that week. The written version sits right alongside for people who would rather read than watch.
• The Nonsense Emailer. My newsletter. The good stuff from the site, a few links I liked, and at least one joke that I thought was funnier than it was. Casual, useful, fun.
• A tools and AI hub. I spend a lot of time testing AI tools, prompts and productivity setups. When something works, I write down how I did it so you can copy my homework.

The five pillars

Everything on the site falls into one of five pillars. Think of them as character classes.

Reviews. Gear, gadgets, audio, smart home stuff. Field-tested, not sponsored. If I say I like a speaker, I lived with it first, and nobody paid me to say so.

Tools and Tech. The nerd stuff. No fear, just upgrades. AI skills, plugins, model reviews, prompt recipes, and productivity stacks. I explain things in plain language. If I use a piece of jargon, I define it on the spot.

Adventures. Travel, music, street culture, dispatches from the road. Expect murals, record shops and at least one food recommendation per post.

Journal. Essays, what is in my bag, and behind the scenes notes like this one. The slower, more personal posts.

The Resistance. Social justice, framed like a video game boss fight. Authoritarianism is the final boss. He has a big health bar and a bad attitude. The party is large, though, and we brought snacks. I care about immigration justice, voting rights, prison reform, a free press, civil rights and putting good tools in everybody's hands. I will keep it funny, and I will keep it serious where it counts.

House rules

A few rules I set for myself. Short sentences. Plain words. Every industry term gets explained the first time it shows up. No sponsored opinions dressed up as reviews. Be kind.

So why "same donkey"? The site is new, and the guy running it is the same one as before. Same curiosity, same love of fantasy novels, same weakness for 8-bit everything. The stall is nicer now, that is all.

Want to keep up? The easiest move is The Nonsense Emailer. There is no fancy signup form yet. You email me, and I add you to the list by hand, like a medieval scribe with Wi-Fi.

Thanks for stopping by. Peace and love.

Somebody asked me recently how a person ends up doing AI work after 25 years of festivals, touring, and software, and I gave the tidy answer, the one that makes a career sound like a plan. The honest version is that I said yes to things that looked slightly beyond me, and the only skill I built consistently was getting comfortable in rooms where I was the least qualified person present.

Here is the long version, with the parts I usually leave out.

The field

I was part of the founding team that built and launched the culturally iconic Electric Forest music festival. That sentence has been in my bio for years and it still does not carry the texture of it, which was mostly a lot of people standing in a field in Michigan arguing about where a path should go.

I booked roughly half the talent and helped design and curate the stages and themed areas. The booking half is the part people ask about. The stages and themed areas are the part I think about. You are building rooms for a person you have never met, who will walk into them at two in the morning, tired and delighted, and decide in about four seconds whether this place is theirs.

It grew to 100,000+ attendees annually. Scale was never the interesting number. The interesting number was how many people picked up trash they did not drop, which turned out to be a lot, because people take care of places they love. You cannot buy that with production budget. You get it by sweating details nobody is billing for.

The agency

At Madison House I led agency growth and operations on the artist management and booking side. A hundred-plus touring artists, which sounds like a roster and functions like a small city with a routing problem.

People imagine this job as phone calls with famous people. It is mostly calendars. It is holding a date in three markets while a support slot gets confirmed, and knowing which promoter returns a call on a Sunday, and having the conversation nobody wants to have at the exact moment it will still do some good.

Artist development is a long game and the arcs look nothing alike. A club run needs completely different scaffolding than a festival slot, and working the Mobb Deep catalog taught me more about how legacy value compounds than any strategy deck I have ever sat through. Later, watching an artist like Thundercat build an audience that came for the musicianship reminded me that patience is a strategy, not a delay.

The very loud night

We sold out Madison Square Garden with Bassnectar.

I have written and deleted three different paragraphs trying to describe standing in that building with the lights down and the room full. What I remember most clearly is the hallway beforehand, holding a laminate, running the same three worries on a loop. Then the doors opened and the worries became irrelevant, which is how it goes every single time and which I have never once managed to believe in advance.

That night is also the clearest proof of something I say to every operator now: the glamorous outcome is made entirely of unglamorous inputs. Advances, settlements, routing, seventeen emails about a load-in door.

The software

Somewhere in the middle of all that, I helped build and license Madmin, an early cloud-based CRM for the music industry. A proprietary booking CRM, built because the tools we had were making us worse at our jobs.

This is where my career quietly turned. I had assumed software was made by a different species of person. Then I sat in requirements meetings and realized the hardest part was not the code, it was knowing which button needed to exist, and I knew that cold because I had needed the button at 1am on a Tuesday with a deal falling apart.

Operators make good product people. If you have felt the pain you can describe it precisely, and precise descriptions of pain are most of what a product team needs.

What the yeses were for

I did not plan any of it. Each step was somebody asking whether I could handle a thing I had not done before, and me deciding that figuring it out in public was better than the alternative.

The through line turned out to be partnership work. Find people doing something good. Get them the room, the budget, and the cover they need. Step back and let it be theirs. That is the same job whether the deliverable is a stage in a forest, a tour routing, a CRM, or an AI system that has to behave itself under pressure.

So now I do AI solutions consulting and agent quality work, and I am writing Coursera courses on deal-making, business, and marketing, which is mostly an attempt to hand over the parts of this that took me a decade to learn and about twenty minutes to explain.

If there is advice buried in here, it is small. Say yes to the thing one size too big. Take notes while you are confused, because the confusion is the most valuable and most perishable thing you own. And be good to the people in the hallway before the doors open. They are the whole job.

Working on something in this territory? Come say hello.