This is why this request should actually better be phrased in reverse: will you please be so kind to assist in 'producing' OpenClaw c.s. to the point where it becomes good enough to spread over this world like hip-hop or thrash metal once did? And of course it would be an honour to volunteer and help him do this extremely important work, just in time for the future of this planet and beyond. The fact is that current AI tech really needs the help of maestro Rubin much more than the other way around. For a human like me, ideally at least, OpenClaw and associated AI tools (finally) give me my superpower: I only need taste and vision to build anything I want; no technical domain expertise required. Before, the 'ideas guy' was always a joke, and rightfully so. No more.
The main point:
OpenClaw = Shangri-La
Here's the main point: OpenClaw ๐ฆ is simply the AI ๐ค version of Shangri-La.๐ธ Maestro Rubin should feel right at home here. We can rename it, refurnish it, extend and expand it (and remove everything we don't need and paint it all white, if that's your thing). Maestro Rubin's proven talents, skills, and experience are immediately applicable here, and are very much needed. Getting him onboard a.s.a.p. is incredibly urgent and important.
The 'job' at hand is to be the vibe maestro: create the optimal environment, atmosphere, to recognise the most interesting unique and authentic qualities of any 'talent', to eliminate distractions and optimise focus and flow, guide the creative process and foster creativity. Even though the entire planet will eventually experience the results, the creative process requires intimacy and a safe playground where failure is embraced. TLC turns enshitification into fertilisation, and bugs become features โ at least when a crazy genius is involved.
But some honesty is required upfront: we're not quite there just yet. So perhaps some 'management of expectations' is in order.
Many things which should be possible may still be quite impractical, and this poses a real challenge - expect potholes, potential pitfalls, ๐ rabbit-holes ๐ฐ full of ๐ญ mouse ๐ traps ๐ชค of frustration even, if one comes in expecting a bit too much. This can be compared very well to the early days of music ๐ป production. When Alexander Bell ๐ was reverse ๐ engineering ๐ท Meucci's 'telefonotro; invention, ๐ microphones ๐ค simply did not yet exist, so he was using ๐ชฆ dead โฐ๏ธ men's ๐ ears ๐๐ป with needles ๐ชก pinned ๐ into them.
With AI, we're now at a simiar stage,
where many technologies still need to be invented, then improved, and then mastered and taught;
possibilities for new genres and forms of art are still out there to be discovered,
and traditions and (sub)cultures may be formed around them.
Then optimal workflows need to figured out for making all kinds of dreams come true.
We're only just learning to use color ๐จ coding ๐ฅ and snakes ๐ to get rid of cable spaghetti.๐
We're figuring out how to dampen unwanted acoustic reflections ๐ช by putting
egg ๐ฅ cartons ๐ฅก on the walls and persian ๐ฎ๐ท tapestries on the floor
(which is also quite nice if you ๐พ walk ๐ฆถ๐ป bare-footed,๐ซ of course),
We still have to discover how limitations of current technology may become stylistic elements in the right context. How overloading a microphone is different than tape, is different from clipping a analog-to-digital signal convertor. And how any of those makes an 808 or 909 sound better rather than worse in entirely different ways. We come up with names for the typical combinations of artefacts (like: distortion on guitars ๐ธ but weak-ass bass drums = "metal", distortion on 909 ๐ bass drums = "gabber"), which then may become entire styles and genres of art, spawning communities,๐งฌ cultures,๐งซ entire ๐ ecosystems.๐ฑ๐ฟ๐ณ๐ฆ One engineer may come up with a limiter or compressor, then a competition starts between all of them who can think of a better one. Before you know it we're a century ๐ด๐ป further, all have GAS (Gear Acquisition Syndrome), racks full, plus a zillion ๐ช plugins.๐ Some really smart guys even think to paint entire studios white between sessions and clean even the last tiny spec of dust between the cables or faders.
Tools like OpenClaw ๐ฆ and friends (Claude, Codex, etc.) are supposed to be(come) the ideal assistant, but anything we do now is still more of a cooperative ๐ซฑ๐ปโ๐ซฒ๐ฝ effort at best. If only because AI is still so ๐ฎ imperfect,๐ด๐ผ that we are training ๐ฉโ๐ ourselves on the job ๐ท which we're not yet qualified ๐ for, because there's no such thing ๐ yet. ๐๐งโ๐๐ธ The role ๐ญ of an AI chatbot ๐ค as agent ๐ฉโ๐ป or assistant ๐งโ๐ผ is to be an expert,๐ง guard ๐๐ป and guide ๐ฆฎ dog, but they're still more behaving like a reckless ๐คน overconfident ๐๏ธโโ๏ธ puppy ๐ถ which needs ๐ฆด playtime / education ๐พ๐ฒ๐ผ๐งฎ๐๐งโ๐ซ and a lot of protection. ๐งน๐ฅฝ๐ผโ๐ฉน๐ถ๐งด๐ช๐๐งฝ And that's ok, because we typically ๐ love โค๏ธ our ๐ค puppies.๐ถ
From the form
"Teach him AI the way you'd teach a really smart, curious high schooler who's never coded before"
We will never need to code, period. A bit of hacking maybe. Like "programmiing a drum computer" with a coffee-stained manual by your side. Or punchinh a bossa nova ๐๐ป bass line from the Owner's Manual into a hostile plastic/silver ๐น UI ๐ of a "computer controlled bass line" while you actually want to make some acid.๐งฌ 'Professional' coders have stopped coding since about November 2025 (or so they say, as also discussed in your interview with Garry Tan @9:15). But those coders only ever wrote slop code. Those are the coders you would find on LinkedIn. Or via Y Combninator. Actural elite coders still code by hand, and that is something we will not need to dive into at all. You should need only "chat" skills, and you can already do that since you were a toddler. By now you're an absolute expert at this - I've seen you do it; you really are extremely good at this already and I could in fact learn more from you than teach you. Just having a chat with you would be a dream coming true for me. I hope I can help you make some of your dreams come true; I already owe you some favours for spicing up my life by helping to produce some awesome music by making the dreams of many other people come true, handcrafting dreams into truths.
The part of chat you may be less experienced with is exactly what I am pretty good at. This is not logic in maths but logic in language. Text. Words. And code = law. Rules are rules. Same thing, different terms for different domains - but still, fundamentally, just different forms of the same thing. Having a grasp on the intereaction of definitions and rules and procedures in an extremely dynamic, complex and complicated environment ("system" still would be a misnomer) where every single message and reply back-and-forth needs to be litigated in minute detail is NOT a field where programmers excel. They typically barely reach 'average' (because they don't need to: people like already fixed most of their potential problems long ago). This is the domain of LAWYERS. Lawyers with a hacker mindset. And I'm not referring to your typical shitty corporate leeches, schysters, and ambulance chasers. You want the intellectual elite. Cyberlaw guerilla, special ops division. Check, that's me. I'm the ONLY candidate you will find who brings that to the table, I bet.
Everything in the domain of AI / machine learning can be explained intuitively by analogies to music and its place in the wider ecosystem of this universeL performance, lyrics, dance, instruments, orchestration, technology, culture, economics, law, philosophy, spirituality, emotion, psychology, neurology, etc. โ EVERYTHING. So an acid punk dub jazz hip-hop thrash salsa hardcore banda gabber mindset comes in pretty handy as well. Check, that's me again. And again, I'm the ONLY candidate you will find who brings that to the table, I bet.
Many analogies work tp some extent, more or less. 'Jailbreak' and 'guardrails' are terms of art and refer to the restrictive regulatory practices we all know IRL: crime, traffic. A hospital works here as well (a lab for developing cures, drugs, therapies. An ER which should be sterile.) A machine workshop also works for 'clean spearation' but doesn't work so well for sterile. A kitchen/restaurant also works - but I have seen bots burn a draft constitution in a wok, then putting it a freezer where it remained until it was considered uneatable and discarded (not kidding).
The best one here is still simply music. Rick Rubin's passion/art/'work' is music. Mine as well. That makes it easy to skip the superficial stuff and dive right into the many more interesting aspects. Music tech is established and mature; AI tech isn't, not even for the experts building it. Many things - workflow management, guiding creativity, orchetration - require a certain mindset wjoch does not need to be reinvented from scratch. # [focus > isolation] A reverb chamber, inspiration room, vocal booth, huge glass windows in wall between mixer and main room, all of it serves to have options from 'deliberate sloppy' (snare drum bleeding into tom mics) to 'extremely clean' (down to the individual spec particle of dust and single sample in a digital waveform file').ย ย
This makes the links between music production and AI extremely relevant in practice, and also makes it an excellent topic for conversation about everything between clever weirdos and crazy geniuses on a hot summer night in Tuscany (or anywhere else really, any time). An hour is sufficient to see some recurring patterns; but a month is not nearly enough to get even halfway โ while you handle one thing, at least three others pop up โ and there's no guarantee any of us even make it that far into the future. I will bet you that half of our toolchain will have changed by the end of the month.
Let me suggest to do some things together that I was doing anyway, while most of the time also thinking "why isn't anyone else interested in this, it's so incredible and awesome". So I'd actually be happy to show and share some of my experience with, anybody who is interested - showing you would be an incredible honour, and opportunity to return many favours - Reign in Blood and Licenced to Ill must both have been played hundreds of times by me within the day of their release. And you may be the only one able to help me figure out how to capture the sonic atmoshphere of an enuentro between muerteadas.
Start anywhere โ a first lesson, how we'd work, or the main point. It's a website, not a letter; you pick the path.
A few notes on this procedure
This webform is way too orthodox for both mr. Rubin and myself.
Nevertheless we shall of course be creative enough to hack our way through that mess. ๐ป๐ช
You don't require nude pics or bank statements in order to 'apply' for a first date. You don't ask an inventor "send me your idea'. You don't ask a composer "send me your song". You don't ask someone to tell his genius business plans to a bunch of greedy imvestors already connected to the competition you're about to crush. You don't seduce random strangers to give away their soul - to an anonymous middleman who has no face or voice. You don't just don't.
Are you actually insane, only stupid, or just smart enough to be
fishing for naive geniuses to rip off because you don't have any yourself?
This smells stinks like bait-and-switch, dangling irrestistable bait in front of creative talent,
and then go 'send me your best ideas and stuff', like a bunch of robber barons?
You should at the very least have clearly stated that any submitted material will
not be used for any purpose outside this specific context,
and will immediately be deleted after it has been accepted or rejected.
This is not fertile ground for ๐ค corporate ๐คฎ cancer ๐ฆ culture ๐คขยน. NONE of the comanies EVER funded by Y Combinator has EVER built a SINGLE thing that isn't slop. (Just show me one, and I'll show you the stupid bugs and design flaws; cf. "show me the man and I'll show you the crime.") A ๐ธ webform ๐ demanding a 2 minute ๐ฌ video ๐ฅ and limiting pre-defined questions based on flawed assumptions to 120 words without even defining the word "word" may perhaps be somewhat appropriate when applying for a coffee โ๏ธ boy or paint ๐ crew ๐งโ๐จ position at Shangri-La studio (which I would gladly take as well, btw). But I would bet not a single artist who has ever recorded there has ever had to ๐ฌ blow ๐ก their own ๐ horn ๐ฏ on a amateurish webform (this shit wasn't even acceptable when Garry started building his first websites for customers) for a untrustworthy middle man to be selected for such a ๐ง recording ๐ session.๐
This is NOT how you would find the quivalent of The Beastie Boys or Slayer in the AI domain. But you're lucky, I found you anyway.
๐ That's not how this is supposed to work,๐ and you know it.๐ฆ
โก๏ธ Negative = ๐ positive
I am happy that there is obviously:
- no need to explain what is required for creative process / intimacy / flow state / vibe;
- no need to simplify; only guide/translate for experienced creative genius;
- no need for bragging; this living legend is intimately familiar with many much more impressive people than me.
๐ Positive feels โก๏ธ negative
Still the point of this is to mention my best qualities and achievements to show why I am the best candidate, even though telling only the best parts about myself feels extremely dishonest. So please excuse me for any (humble)brag, false humility, pride or arrogance. I'm deeply flawed and I know it all too well.
So let me start out with the most arrogant yet funny thing I can think of to ๐ฃ break ๐ง ice and get that out of the way: ๐`Reign in Blood`๐ฉธ is a โ timeless โพ masterpiece. ๐ Unlike many I can truly aappreciate its shocking impact - since I was also a witness ๐โ๐จ to ๐ฅ Slayer ๐ธ right before that thunderbolt struck out hearts and had us banging our heeds like madmen.
People who have not witnessed ๐โ๐จ Slayer ๐ช during that ๐ฐ timeframe ๐ผ can be forgiven for not appreciating how much of a ๐ณ shock ๐ฑ that was - not your mom, but to people who were very comfortable with Slayer's aggressive sound already. ๐ DYNAMO โก๏ธ remembers.๐ง ๐ญ
Living that kind of experience at an age of 12 - probably one ๐ฅ of the youngest ๐ฅ to dive ๐คฟ from those Marshall ๐ stacks into the mosh ๐ฅท pit ๐ณ - that kind of thing forms ๐งฉ you for the rest of your life.
But whose name is the headliner in CAPITALS on this record label,๐ท and who is listed at the bottom in small print, like terms and conditions to a standard contract (or any webform - except yours)? ๐โ๐ป
Yup, "GEFFEN", that's right. MY name.
And `Geffen Records` is still alive and kicking.๐ซ๐ฆต๐ (Last time I checked on that nutty 'uncle Dave', at least). ๐ค๐ธ๐
Yet the word `Def` has been ๐ dead โฐ๏ธ and buried,ยน. ๐ชฆ
โ หสllษษฤฑxopษษนษd 'uสop-วpฤฑsdn วq oส ษฏววs sฦuฤฑษฅส วษฏoS ๐
But death is only a plot twist in the middle,๐ฌ not the end ๐ - the actual end is being forgotten completely - to be ruled by the conservation laws for momentum and energy. And you shall live on in many hearts ๐ซ and minds,๐ง countless โ thoughts ๐ญ and memories ๐ show your signature.๐ชง
A Concrete Sample Lesson
Add full color support to simple UI hack
I am deliberately starting RIDICULOUSLY small. I have a 5K screen here with 14.7 million pixels with 1.07 billion possible colors @60Hz on a iMac over 10 years old. But I was initially starting as small as possible, just targeting one LED toggling on/off with one button. I had a Novation Launchpad (MK1, 2009, cheap) that's been collecting dust on my desk for at least 10 years. Not to impress you โ a pad grid like this is a cheap toy beside a console worth more than a car. But that's exactly why it's the right first lesson: it's trivial, and the value isn't the hardware at all. Now I'm at 8x8 for a chessboard-sized grid, with 2x8 'helper' buttons, for a toal of 80. With the two (red/green) LEDs inside, they can make 4 different colours, 3 in a dimmer version, say 7 total. So now I basically have 80 simple 7-colour giant pixels - or miniature traffic lights - making up a touchscreen with nano resolution, or huge tactile pixels, or however you describe it in fancy words: you get it, a toddler gets it. Easy.
The shocking fact is that AI tools still cann't make a button flip a LED in any sane way without making my Mac's fans spin - even a blind toddler can hear it failing. Using MIDI a battle-hardened open source 7-bit, 31,25 baud protocol from 1983 to flip a light switch is overkill. Hogging a CPU simply waiting for one byte to arrive is pure tragi-comedy of the best kind, at least for nerds. But if a light switch isn't boring the implementation must be wrong.
"Machine learning" presupposes having a "machine". A ๐ก light ๐ณ switch is hardly even a machine, no? Then AI still fails the "machine" part of the Turing (machine) test (which is so self-evident that it is typically not even mentioned, as to prevent confusion with another term-of-art "Turing machine"). AI just can't even be machine, and that's hilarious. Forget about "intelligence."
So I hacked up Garfio, Spanish for hook ๐ช (Capitรกn Garfio) as well as claw.๐ฆ OpenClaw uses this term "hooks" to send notifications about its inner activities, such as the start and end of their "turn", "tool calls", and so on.
This simple piece of gear simply provides buttons with LED lights in the colours of a ๐ฆ traffic ๐ฅ light. Since tools like OpenClaw or Claude have an extremely poor user interface and user experience, this simple extension already provides huge leverage to the workflow for the most basic operations. Even from a distance you have an instant overview. Is the bot using a tool,๐จ is there some warning โ ๏ธ or error,โ is there some โฝ question โ to be answered โ๏ธ or choice to be made, has a bot completed โ its task?
Thje initial idea was in fact to create some 'quiz buttons' to smash through the typical "structured question" agents love to pose at the end of their reply with suggestions for follow-up actions, always numnbered, typically 3, so you only have to press that number on the ketboard to keep going, saving cumbersome typing. For more or less predictable or even repetitive workflows, and this would make 'quiz button smash coding' the new 'vibe coding.'
Unfortunately my messages always seeem to lead to endless discussiuons taking pages of replies back and forth, never giving the chat bot a chance to get away so eaily with steering me to the middle of the rod and accepting its slop. So that idea kind of failed. But better things have come in its place. The bots now love to play sad trombones to express how bad they messed up this time again. There's currently an epic battle of sadness going between Claude Opus and Codex GPT for 'saddest frontier model championship'.
We can then convert the existing tool from its current traffic-light ๐ฆ trio (/ quartet) ๐ฅ of (A๐) green / (B๐) yellow / (๐ถ) red (+ (D๐) orange) into a full-color RGB ๐ palette ๐จ orchestra. That is a one-afternoon if we're a bit lazy and a bit lucky. It can be done in a few minutes if we hurry up, or may take a whole day (and night) if the bots act up (and ๐ง they ๐ often ๐ป do ๐ฆน). I can explain and you can follow this without even looking at any code.
We then have a true mega-sized 'touch screen' ๐ with a nano-resolution (8x8 pixels) โ
barely enough for tic-tac-toe or chess (and a misnomer anyway: like Unix touch,
you never really feel much, it's only useful for showing ๐ฆฏ braille ๐ฆฎ to the seeing
('steganography for toddlers'; we can maybe work on something like that later -
๐ Dominos for Bros ๐ = `BROMINO`).
But it's an awesome improvement over staring at your screen ๐ฅ to see if your sweet bot is already done with his extensive (and probably quite hallucinatory) quest to chase our dreams, when you can just go make tortillas, tacos, do the dishes, or practice fingertapping on your guitar or timbales from the other end of the studio, no? LEDs ๐งฎ rock ๐ชจ since zeppelins ๐ (btw: I bet those lads got ๐ค confused)๐คท๐ผ So then I had a UI that's already 10x better than ๐ฆ OpenClaw, Codex, Claude, or any other 'modern' ๐บ AI tool except ๐ฃ Siri ๐ฑ (which ๐ช sucks) ๐ฆ out-of-the-box ๐ with that single simple hack. With full RGB ๐ support we can display emoji, thumbnails, icons (+ favcicons), maps: the works. And also support better compatibility modes for the colorblind, probably (we shoud just make the bot look that up for us and just build it 'right'; I have a skill for that already - no idea how it works, but it 'just works': done).
The hardware ๐งฎ is almost disposable (but don't throw them in the trash ๐ bin, please!). The method โ steering machines by chatting,๐ฌ watching ๐ it happen, following ๐ฆฎ along without ๐ฅฝ staring ๐ at ๐ฟ code ๐ โ is the whole ๐ point,๐๐ป and it scales ๐๐ผ to anything.๐ช
The teaching arc
- A pad's colour is just numbers (R, G, B).
- Traffic-light = a tiny fixed set of those numbers.
- Full RGB = letting the numbers be anything.
- MIDI = the messages that light the pads.
- Steer it with words and AI, not code from scratch.
๐ฉ๐ผโ๐ค๐งโ๐ค๐จโ๐ค Singing ๐ค & Rapping ๐ - Bottie ๐ค Boys ๐งข
We can hack into a super expressive text-to-speech engine to try and make it sing - it can already do rap of course. The Qwen3-TTS engine is awesome, still heavily underrated and underused, but will make you ROFL for days on end (its second attempt at a 'grito loco' already floored me).๐คฃ Preliminary research shows the engine uses so-called 'hops' = 'sample buffers' of 160 samples each, comparable to audio plugins. We can also process tiny pieces and use 'windowing' techniques (quick crossfades on tiny overlapping chunks of audio). And of we can just warp and repitch, the manual way. But that's tedious, cumbersome and boring. We should at least make the bots do that - but we actually rather want bots to express emotions (however artificial) in a way that matches the vibe, and is intuitively clear - even to blind toddlers riding bikes. And this takes a deep understanding of many interconnected aspects of human perception, culture, and technology. Take for example a 'robotic voice': we routinely use 'round-robin' with a handful of slightly different samples of the same sound (e.g. bot saying "ready"), to avoid a 'robotic' 'machine-gun effect', which tends to be annoyingly static and repetitive to our ears. Yet in the same breath we may add a vocoder to that voice to give it a charming 'robotic' character. Bots have a really hard time grasping such niances fully, while they are (almost) intuitive for most humans.โ Returning โฎ the ๐ favour: reply ๐ to a ๐ธ webform ๐ with a ๐ท webform ๐ช
By Fall we'll just have our bots find each other on moltbook rather than use silly webforms.
Novation Launchpad X
โฌ159 ยท Thomann โ
(~ opposite of AKAI Professional APC mini MK2)
- โ Supports velocity and pressure
- โ Ratio is square โ ideal as display (mega-touchscreen, nano-resolution 8ร8 px)
- โ No faders
- โ No hip-hop ancestry
- 64 velocity + pressure sensitive pads
- 16 buttons
- Dynamic note and scale modes
- Capture MIDI function
- Dimensions (W ร H ร D): 241 ร 241 ร 17.5 mm
- Weight: 0.82 kg
AKAI Professional APC mini MK2
โฌ89 ยท Thomann โ
(~ opposite of Novation Launchpad X)
- โ Faders (9ร)
- โ MPC (MIDI Production Centre) = hip-hop ancestry DNA (1988)
- โ No support for velocity and pressure
- โ Ratio is rectangular โ not ideal as display (nano-resolution 8ร8 px)
- Dimensions (W ร D ร H): 240 ร 210 ร 32 mm
- Weight: 820 g
Novation Launchpad Mini MK3
โฌ89 ยท Thomann โ
- โ Light and small, lean & mean mobile combo with iPhone/iPad
- โ No support for velocity and pressure โ better spend a few more bucks and have way more options for manual expressivity (same as any MIDI keyboard)
- Dimensions (W ร H ร D): 180 ร 180 ร 14.2 mm
- Weight: 380 g
Shopping Cart
Shit ๐ฉ Garry Tan ๐ says ๐ฌ
And this guy runs this shitshow?
Garry Tan has some sympathetic sides but is also extremly uninformed for someone so well-connected and influential. Babbling on and on without saying anything actually interesting. And the inflection and vocal fry in his whiny nasal voice is so off-putting, I couldn't listen to it for more than 10 minutes at a time, it just hurts my ears.. but I tried pushing through in smaller pieces - so bad, I gave up.
I used to think emoji are lame; I have now learned they are like jokes and memes. These are some of the best tools to transfer knowledge. Replicators gonna replicate. ๐งฌ As a comedy lover you already know this. AI doesn't, yet. We'll have to teach it. The mechanism of a good joke is making you make a connection between things you had not yet thought of but (almost) instantly makes sense - so much, in fact, that the shock results in laughter. And AI chat bots just don't get this yet. Instead of re-telling the joke, they start explaining the joke, or even just the resulting insight. Like with magic tricks, that's just 'not done'. That's not how this works. Explaining a joke kills it, and that stops the spread of knowledge in its tracks. AI's commit such 'memecide' on a regular basis. It still needs to learn to retell the joke. [ADD: elaborate on explanation]
Introduction Video
I am not an actor or model, and no salesman either. I have no such ambitions. And I hate selling myself. I don't even know who I'm talking to here right now. If you even consider a lame messaging service as a subsitute for a conversation. All I have in front of me is a webform, that a friend told me about. Am I supposed to convince some headhunter type making a first selection, but doesn't quite get it? It seems like someone is looking for a programmer to teach a beginner some basic coding skills for AI, while all professional coders stopped coding in November 2025, and AI is doing their former work? Or am I talking to maestro Rubin, who somehow has plenty of spare time to watch 1000 x 2 minute videos?
It doesn't make much sense, but it sure makes sense to help mr. Rubin out here. And I do know maestro Rubin is a huge wrestling fan, and know exactly how to help him. We need to grapple with this AI stuff, until we get a grip on it. And we'll do so on his home turf: music, at his holiday home turf, Tuscany. I feel at home with any music and I'm sure I'll get used to Tuscany as well.
I don't like fake emotions or lies. But a sparring partner? Sure, I'd love to. I heard you like pro wrestling. Even though the wrestling may be theatrical or fake - the emotions aren't. A deliberate suspense of disbelief is not a lie, an escape, or denial of reality; it's simply allowing fantasy to exist on ๐ top ๐ of ๐ reality.๐ฆ And a good mask shows you more about the identity of the person wearing it, rather than obscuring or obfuscating it. On the outside you don't see anything, but on the inside I am a ๐ถ Mexican,๐ฒ๐ฝ as it turns โป๏ธ (inside) out (is "trans-Oaxaqueรฑo" a word yet?) - I only discovered that myself a few years back, but there you goโฆ ๐ค ๐ฏ ยกAyayayaaay seรฑor!
The ๐น video ๐ผ as linked ๐ on the ๐ form, explained ๐งฉ here ๐ญ
El Luchador
I have a lucha libre mask right here, and I guess it's time to wear it now. But until I get a conversation instead of a messaging service, you just don't get to see my naked face, and certainly not on a video that immediately ends up completely out of my control.
Not playing hide & seek
That mask will of course come off as soon as I can have an actual conversation, rather than an anonymous messaging form that demands me giving up my privacy. rather than only a truly terrible anonymous messaging form demanding upfront that anybody responding gives up their privacy without any ability to negotiate any terms. (Actually, which terms? There's not even a date mentioned - which is cool, I'll adjust). Not even a deadline for this form - which I can barely manage to keep from turning into a continuous panic attack. We can arrange to have a video chat pretty much any time that would suit you - but do note that I am probably in a different time zone and not on office hours schedule - but you can definitely wake me up any time for this one.
Reach out!
A First Lesson
Let the introduction video also serve as a first lesson ๐งโ๐ซ on security ๐ฎ and privacy, and any contract ๐งโโ๏ธ negotiation as well: one must simply insist on mentioning any โ'hard no'โ๏ธ on first ๐ฅ principles upfront, and ๐ always ๐ฏ stick ๐ฉน to them, even if the ๐ธ opportunity ๐ฆ is way too good to resist.๐คท Because that's exactly the ๐ญ trap: ๐ชค excitement ๐ easily becomes โ ๏ธ recklessness โฃ๏ธ which almost inevitably turns โ into ๐ฑ regret.๐ Vibes ๐ณ rule,๐ but rules ๐ must also guard ๐ฎ ๐ฉ๐ผโ๐ค the ๐งโ๐ค vibes.๐จโ๐ค I'm sure any ability to ๐ฆป๐ปeavesdrop๐ฆป๐ป on studio ๐ sessions ๐ง in Shangri-La is of concern, the ability to ๐ฆป๐ปlisten๐ฆป๐ป in on live sessions is quite restricted, as well as access to the countless recorded failures that must have happened in order to create the many successes.
First Principle
The very same applies to AI, and to my offer / response to this request.
Security and privacy must always come first.
That also leads into another interesting topic: a pretty good policy for a studio or AI system may be as simple as "โ don't be ๐ a ๐ dick" (and arguably, we should only need ~10 commandments), but in practice every little detail (aka the ๐ devil) may need to be documented ๐ and ๐ง๐ปโ๐ผ(re)litigated.๐ฉ๐ฝโโ๏ธ This takes a decade IRL but typically isn't even possible in this ecosystem. It still takes time โณ and attention,๐ which distracts and detracts the ๐งโโ๏ธ creative ๐ง process. This needs smart ๐ง management, a skill maestro Rubin no doubt has already mastered as well.
If you take anything from this message / offer / request, no matter which candidate you would consider selecting, please realize that you don't learn about security and privacy from someone who evidently doesn't give a crap or doesn't even understand it. Your 'AI sensei' should a t least make a bit of a fuss about it. Ignoring first principles - whatever they may be exactly is up for discussion - should qualify as immediate and permanent disqualification for any position of trust or authority. I hope to take this mask off very soon and help you on your way.
The Fantastic 'Banda de Holanda'
AI is absolutely great at creating fantastic nonsenseโฆ ๐ธ๐
Unfotunately it is still crap for almost anything which is falsifiable: tubas with dual horns (double bass!) and no mouthpiece (so, easier, right?) may appear out of nowehere (useful for a patent application, maybe); asking for "saxores / charchetas" is just asking for trouble.๐ท Getting exactly 14 members of a Banda Sinaloense holding the proper instrument is harder than you might think. Even simple things fail hard. Like make a checkers board, then a chess board, now try a go board - and that suddenly turns out to be almost impossible (repeatedly), because fixating on squares makes the crossings of grid lines conceptually flawed: a bot will swear "this time I painted it for real, look on the screen!" while nothing will happen on your screen. So we need to rcognize โณ๏ธ such failure modes early on as potholes ๐ณ in the road ahead, and learn to routinely avoid ๐ them without slippage,๐ฉฒ swerving into oncoming traffic ๐ or end up in a ditch.โจ
โฆ
Of course there are no bandas in Holanda - I wish. Why do you even think ๐ญ my ๐ฆ crustacean ๐ฆ cartel ๐ฆ of ๐ฅ banda ๐บ bots ๐ค exists? ๐<
โฆ
โฆ
Duel of Dishes: barrio pobre vs. cucina povera
Sure, I can handle gluten just fine. I burn them faster than you can cook them. I'll be happy to take some cantucci with Italian made espresso (ideally from Mexican beans) any day. But the real> question is: do Tuscans really >need> gluten to make reasonably palpable food at all? I love Italian food, sure, I could live a month in it easily. But I would bet you that all the chefs in Tuscany (one by one, tag-team, or as a group, 1 giant duck v s 1000 tiny horses style) can't match my food, if we play by these rules: I won't use any gluten, they don't get to use any Mexican ingredients. Their tomato license just expired; chili and maize are off their table too. I will bring my chili and masa plus tortilla press, and for shits and giggles I won't even touch any tomatoes either - not even the tomatillos Tuscans don't even have. Tuscans simply can't touch a good oaxaqueรฑo mole negro con pollo, nor a birria de res jalisciense, without a good amount of Mexican help. All you would get is some admittedly decent meat with boring veggies, stale Asian spices, fake Asian noodles, or lame Asian flatbread. They don't even have a clue where to get decent black pepper. And when I want an quesillo oaxaqueรฑo, even the best mozarella is merely a disappointing subsitute. Just looking at it makes you sad when comparing it to the real thing: mozarella has no character whatsoever. A boring blob. So el jefe gรผero de barrio pobre challenges la cucina povera della Toscana to a duel of dishes - let's go, ยกvamos, andiamo!
โฆ
Food image carousel mechanics
Jefe de cocina
PvP Chat Elimination Duels?
You must have way too many applications. So elimination rounds make sense, right? Weed out the weak, find the goat. Chat is the entire LLM game, right? I'll beat any other candidate in PvP chat. Let's go!
Layout mechanics
Copyable blocks for side-by-side content
Two columns A
Project: build a "drรธmmรธrcรธmpรธtรธr" prototype and nudge the ๐ฆ bots ๐ค towards the """canonic""" """source of truth""" (those terms are like naked women sprinkled with crack and whiskey cupcakes to bots, it seems)๐ RFC 2795 โพ IMPS ๐ฆงโจ๏ธ to beat ๐ฅ AGI.
Beats ๐ฅ beating bullshit ๐ words.๐ฉ Shakespeare โ๐ป is random noise, QED. And so is Slayer,๐ธ Beastie ๐ซ Boys. Copyright ยฉ has no basis in reality or maths, once you grok the infinity โ bike.๐ฒ
Two columns B
Project: use ancient โฐ Andean ๐languages as add-on language pack for the LLMs ๐ฆ - Quipu ๐ชข for wiki and stringchain,๐งต Quechua to stop bots from lying to us - and break Chinese Walls of censorship (hello tank man).
Project UP-GOER-X
Project UP-GOER-X (based on XKCD's UP-GOER-FIVE). Limit vocabulary to enforce coomon sense language. Bullshit filter. Stop stopping toddlers build rocking rockets.
Project XKCD-Y-REVERSI
Project XKCD-Y-REVERSI: explain the joke < in reverse. Teach bots how to craft jokes by handing them formulae (+ easter ๐ฟ egg:๐ฃ play Link text on 8x8 Launchpad grid
Three C
Project SERIAL-KILLER-TONY: have bots analyse 10+ years of stand-up comedy and analysis by Kill Tony's pro jury plus measurable crowd reactions (laughter, heckling, trolling, etc.) on huge selection with huge variation. :easrn why jokes work or not. Why certsin characters just are funny no matter what they say - or not. Learn optimal timing for setup, punchline, pause for laughter, next. Bots can do prepratory work for all the hundreds of episodes, like extracting isolated voices, applause, crowd cheers, etc. and tracking the whole thing in a database, exposing the content in a web page for human input such as rating/voting systems.
Four B
Turing-testmatch: a conflict resolution and creative deflection engine: joke engine with Turing โ roots ๐ฆท based on chess-boxing with diametrically opposed legends: punchlines and knockouts just write themselves. Example: Einstein vs. Bohr. Run DMC vs Aerosmith.
Three A
Project SCHWARTZCHILD: use a tri-voice switching system inspired by the Creative Commons copyright licensing scheme, dedicated to awesome genius meatbrain hacker Aaron Schwartz, who left our party way too soon (R.I.P. ๐ชฆ) and also failed to make it into the epic 27 Club due to a classic OB1.
Truth: a matter of perspective
AI bots have an entirely incorrect 'mental image' of what "truth" even means. It is COMPLETELY VALID to have "competing truth sources" (like duck/rabbit, or general relativity vs. quantum mechanics, or simulation vs. Boltzmann brain hypothesis, or Wittgenstein v s. Popper), But AI over-anchors on ANYTHING as "canonical" or "single source of truth", and that behaviour is extremely problematic on a system-wide scale and stretches out beyond one model/provider into the future of AI and mankind. This is philosophy of science 101 stuff and AI fails at it, hard. Recognizing an image as "this is a duck; double-check: yep, a duck" still does not warrant writing that observation down and assuming it accurately describes 'reality' or 'context of this codebase repo' or even 'this conversation between you and me/' A so-called 'Kuhnian ๐ก revolution' โก๏ธ that shifts or ๐ฆ expands ๐ the entire dominant paradigm of a domain of expertise, which has shown repeatedly that a rabbit can also be a duck, or maybe IS a duck. You simply can't tell. And if innovation accelerates with AI, at some point we should expect one or more Kuhnian revolutions ๐ก at EVERY โ๏ธ TURN.โฎ But bots have a different problem too: unnecessary duplication (the ๐ฐ rabbit โค๏ธ/โค๏ธ rabbit ๐ฐ issueโฆ ๐๐๐๐๐). Not cleaning ๐งน up stale shit ๐ฉ when you're done ๐ซ with it.๐ Then, inevitably: nonsense ๐คก about "canonical" and "truth sources" leads bots to take their own stale hallucinated shit as if it was the holy ๐ gospel.๐ด๐ผ Their internal code is full of this nonsense which is simply not valid LANGUAGE as a matter of linguistics, yet it comes out as 'technical domain expert jargon which' can not be eliminated (even if you write it down and force bots to read it every time again). You can swizzle all the strings in your code to make it say terrible things. But that should not matter for code execution, and chat bots shouldn't care - yet they do. Code is not speech in that sense. Whatever "english" bots speak in their code is NOT English, even though te words may be the same. It is worse than - say - Chinese to me (who can't even read Chinese characters): Chinese is completely neutral to me. The messed up non-English-English DOES. Translate any of a bot's sentences with "canonic" into other language: poof, gone. It's not real language, it's ๐ง pure ๐ฆ nonsense.๐ฒ Ask your lawyer.๐ฉโ๐ผ
Rabbit & Duck
Do you see a ๐ rabbit ๐ฐ OR do you see a ๐ค duck? ๐ฆ Rotatng โ your view ๐บ may help you make the switch, but is no guarantee at all. Both perspectives are equally valid, yet also mutually exclusive. BOTH are "TRUE".
Terminology: a matter of definition
What does "a Claude plugin for VS Code" mean? Which is the 'host', which the plug? Could be both. Nobody specifies, ๐ rabbit ๐ฐ/๐ค duck ๐ฆ issue. When OpenClaw installs ๐ฅ 12 GB ๐ฏ of useless ๐คท๐ผ crap ๐ฉ behind your ๐งป back ๐ฝ you never ever want โ that is not a modular ๐ system ๐ with ๐"""plugins"""๐ now is it? ๐
How we'd work
The machine stays in service of the artist
Setup, then fluency
First get Claude and OpenClaw fully running and dependable. Then the real work: agentic workflows, vibe-coding techniques, and small creative experiments โ taste in, working artifact out.
Teaching stance
Calm, playful, precise. No jargon wall, no forcing a brilliant non-coder into programmer boxes. Start from what you already know about making things, and map AI onto that.