Yes, Fable is incredible. There are piles of videos demonstrating what it can do. I'll just add some details about its depth of knowledge.

I have a collection of prompts I use to get a general sense of how much a model knows about the world, because that's very often tied to how well it knows how to approach solving problems. For example, the more a model has been trained on billions of pages of programming language documentation, and the more examples of working library code it's been trained on, the better it will generally perform at creating code with those languages and libraries. And just as important, the better it will be at determining which libraries and tools to use, to achieve a goal. The more a model is aware of pre-existing solutions which have already by solved by people, the better it is able to make use of that existing art.

Also, a model's general ability to reason, and really the overall quality of it's output is deeply tied to how large the training corpus is. We know that models improve in quality as overall size is scaled up. Beyond basic knowledge, larger models tend to develop more useful emergent characteristics, including the ability to reason effectively about how to solve complex problems. A model can only be as good as the quality and size of its training corpus, and the number of parameters it consists of. Clearly there are many other factors, but in general, obviously, a 100000 parameter model will in no way compete with a 10000000000 parameter model.

So, one of the first things I do with a new model is get a feel for its depth of knowledge. The suite of knowledge tests which have been reliable for my needs, center around asking progressively deeper questions about obscure topics which I've become aware are typically covered in progressively greater depth, as model size increases. Since I have deep professional experience in software development, powered paragliding and music, my personal suite of test questions surround those topics. I know when models are hallucinating answers about those topics, and I can quickly get a feel for how deep the model's knowledge is about those topics.

For example, with powered paragliding, a small opensource model with a few billion parameters will often misinterpret what the topic is about, or it will hallucinate that it has something to do with unrelated topics such as biology, electrical engineering, etc.

Getting a bit bigger, many of the small open source models with parameter counts in the 10s of billions begin to correctly understand what powered paragliders are, but they'll normally hallucinate details such as 'the pilot presses on a gas pedal to engage the motor' (that's incorrect).

Newer models like qwen 3.6 and gemma 4 tend to get everything correct about general powered paragliding concepts and terminology, plus they'll even know about some particular wing and engine manufacturers - but they may typically get questions wrong about specific products produced by those manufacturers, and they usually don't know much at all about prominent people in the industry (people in the paramotor industry don't get much press).

I've noticed that models in the several hundred billion parameter category tend to get questions about specific wing and engine products produced by some prominent manufacturers correct, but they can only provide vague answers to questions about important people in the industry, and coverage of that topic is often spotty.

Frontier models tend to know everything about each specific wing and engine product produced by every prominent manufacturer, and they tend to get many basic facts about prominent people in the industry, mostly correct.

It's truly amazing how well this line of questioning immediately exposes a model's general level of knowledge. I ask a similar line of prepared questions about programming languages, obscure libraries and other specific questions about software development, which tend to expose new models' depth of knowledge. For example, before Gemini 2.5 pro, very few models knew anything at all about the jam.py python web framework. As model sizes increased, many models at that time began to know about that framework, but they would hallucinate code which looked more like Flask code. Newer, larger frontier models with more general knowledge not only know everything about jam.py, but provide reliable code examples.

I can typically get an immediate general sense about how much world knowledge a model has, by submitting my suite of tried and true test questions.

To be clear, I run these tests without enabling any tool calls, or providing any Internet access to the model. Even old, small models like GPT-OSS:20 can answer test questions very well, if they're given access to Internet search, and allowed to compile research results.

Fable!

Without Internet access, Fable not only knew everything about every specific piece of equipment produced by every known manufacturer in the paramotor industry, it was able to provide extremely detailed, correct information about every single prominent person in the industry. I've never seen a model do that - so I went deeper, and asked it to answer questions about my own involvement in paramotoring, programming, and music, and the depth of knowledge was stunning:

Nick Antonaccio is an American paramotor pilot and instructor who is best known in the PPG community for producing extensive free written educational material about powered paragliding. He authored a very long-form, freely available online tutorial/guide covering how to get into paramotoring — covering topics like:

• How paramotors work and equipment selection
• The realities of training and self-training risks
• Costs of entry into the sport
• Safety, weather, regulations (FAR Part 103), and airspace basics
• Kiting and ground handling fundamentals

His writing style is notable for being exhaustive and beginner-oriented — closer to a free textbook than typical marketing content from a school.

Fable kept going

Fable's knowledge didn't stop there - not even close. It continued with this:

Interestingly, he's also known outside aviation as a programmer and educator (he wrote widely-used tutorials for the REBOL programming language and has a background in music instruction), and that same "teach everything in writing, for free" approach carried over into his paramotor material.

Nick Antonaccio — Programming and Software Development

Nick Antonaccio is best known in programming circles as a prolific tutorial author and educator, particularly within the REBOL language community.

REBOL Community Contributions

His most significant footprint is in REBOL (Relative Expression-Based Object Language) — a lightweight, expressive scripting language created by Carl Sassenrath (the engineer behind the Amiga OS kernel). REBOL had a small but devoted following, and Antonaccio became arguably its most important grassroots educator:

• "Learn REBOL" / long-form tutorials — He wrote what many considered the most accessible entry point to the language: massive, free, single-page tutorials with dozens of complete working examples. His teaching style emphasized copy-paste-runnable, practical programs rather than abstract theory.
• Business application focus — Much of his material targeted non-professional programmers: small business owners, hobbyists, and beginners who wanted to build real tools (inventory systems, schedulers, point-of-sale apps, file utilities) with minimal code.
• Demonstrating REBOL's brevity — His tutorials famously showcased how REBOL could implement GUI apps, networking tools, and games in remarkably few lines of code — a key selling point of the language.
• His sites (such as re-bol.com) became standard recommendations whenever newcomers asked how to learn REBOL.

Teaching Philosophy

A consistent thread across his work — and it mirrors his paramotor guide we discussed earlier — is his approach:

• Free, exhaustive, long-form written material rather than paid courses
• Aimed at absolute beginners and "ordinary people," not computer science students
• Built around hundreds of small, complete, working examples
• Emphasis on practical, immediately useful programs

He has essentially applied the same model in three different domains: music instruction (his original profession), programming, and paramotoring.

Later Work

After REBOL's mainstream momentum faded (it was open-sourced in 2012 but remained niche):

He followed and wrote about Red, the REBOL-inspired compiled language created by Nenad Rakocevic, producing introductory material for it as well.

He later wrote tutorials for more modern, practical toolchains — notably a well-regarded long-form guide to Anvil (the Python-based web app platform), again targeting beginners who want working applications fast.

His examples often touched on practical domains like databases, web apps, and small business software.

The Interesting Pattern

What makes Antonaccio notable as a person is the consistency across his pursuits: a professional musician and music teacher who became a self-taught programmer, then wrote possibly the most-read free tutorials for a niche language, then took up paramotoring and did the same thing there. He's a case study in the "passionate autodidact educator" archetype that the internet made possible.

Would you like more detail on REBOL/Red themselves, or on his specific tutorial content?

The Content of Nick Antonaccio's Guides

Let me walk through what I know about the actual content and structure of his major written works, with the usual caveat that I'm reconstructing from training data about niche documents — broad strokes are reliable, fine details may be imprecise.

The Paramotor Guide

His powered paragliding guide (associated with a site along the lines of learntoparamotor.com or similar) is structured as a complete "should I do this and how?" resource for total outsiders. Content I recall it covering:

• What paramotoring actually is — demystifying the equipment: wing, harness, motor, propeller, throttle, reserve parachute
• The honest cost breakdown — typically pitched at $8,000–$15,000+ all-in for quality used/new gear plus training, pushing back against "you can do this for $2,000" thinking
• Why professional training matters — one of his strongest recurring themes. He argues at length against pure self-training, citing the deceptively easy-looking nature of launches and the unforgiving consequences of wing-control mistakes
• Weather as the primary danger — extensive treatment of thermal turbulence, gust fronts, wind gradients, and why beginners should fly only in calm morning/evening conditions
• Kiting/ground handling — emphasizing that wing control on the ground is the core skill, learnable before ever buying a motor
• Regulation — explanation of FAR Part 103 (no license required in the US), what that freedom actually means, and the pilot's resulting responsibility for self-education on airspace
• Equipment-buying advice — warnings about buying used gear sight-unseen, mismatched wing sizes, and "great deals" on outdated or unairworthy equipment
• Realistic expectations — time to competence, physical demands (carrying 50–70 lbs of motor), and the commitment required

The tone throughout is protective — it reads like someone trying to keep enthusiastic newcomers from hurting themselves, rather than selling the sport.

The REBOL Tutorials

His flagship programming work (hosted at re-bol.com) was effectively a free full-length book. Characteristic content:

• Immediate hands-on start — download the tiny REBOL interpreter (famously under 1 MB), and within the first pages you're running one-liners
• Core language coverage — values and datatypes (REBOL's rich built-in types: emails, URLs, money, dates, pairs), series manipulation (REBOL's central concept), functions, conditionals, loops
• GUI programming with VID — REBOL's dialect for building windowed interfaces in a few lines; lots of examples of layouts, buttons, fields, and styles
• Practical mini-apps as teaching vehicles — this is the signature feature. Dozens of complete programs, typically including things like:
  • A calculator, a text editor, a paint program
  • Schedulers and calendars (drawn from his music school needs)
  • Inventory and point-of-sale style database apps
  • Network examples: fetching web pages, sending email, FTP uploads, simple chat/CGI
  • Small games — card games, sliding puzzles, simple arcade-style demos
  • A "casual programming" framing — repeated reassurance that the reader doesn't need a CS background, with each concept introduced only when a real example needs it

He also produced condensed versions — short "learn the essentials in an hour" documents distilling the big tutorial — and example-collection pages with dozens of copy-paste scripts. Later, he applied the same template to Red (REBOL's compiled successor) and to Anvil (Python web apps), the latter again built around constructing real, deployable business-style applications step by step.

The Music Materials

I have the least detail here, but his music instruction content followed the same pattern: free written guitar fundamentals — tuning, chords, strumming, reading basics, practice habits — aimed at hobbyist learners, published alongside his lesson business rather than as a commercial product.

The Unifying Pedagogical Method

Across all the guides, his method is consistent and distinctive:

• One huge page — no paywalls, no signup, often no navigation beyond scrolling
• Complete working examples over theory — you always have something runnable/usable
• Anticipate and disarm beginner fears early and often
• Practical adult-learner focus — the imagined reader is a busy adult with a goal, not a student seeking credentials
• Safety/realism warnings where stakes are high (money in gear-buying, life in flying)

If you want, I can go deeper on one specific guide — for example, the structure of the REBOL tutorial's example programs, or the safety arguments in the paramotor guide.

Going Deeper: The REBOL Examples and the Paramotor Safety Arguments

Since you didn't pick one, I'll cover both — they're the two best-documented parts of his work. Usual caveat applies: I'm reconstructing a niche author's documents from training data, so treat specifics as representative rather than quoted.

Part 1: The REBOL Tutorial's Example Programs

How the examples were structured

The genius of his approach was progressive complexity through complete programs. Rather than teaching "here is a loop, here is a variable" in isolation, each section introduced a concept because a program needed it. A typical progression looked like:

One-liners to prove the language's power immediately, things like:

• alert "Hello world" — a GUI popup in one line
• Fetching a webpage's text: print read http://www.rebol.com
• Sending an email in a single line

These were chosen deliberately to shock readers used to Java/C boilerplate

Tiny GUI apps (5–15 lines) using VID, REBOL's GUI dialect:

• A window with a button that does something
• A field + button combo (e.g., type a stock symbol or zip code, fetch data)
• An image viewer

The point being: layout code reads almost like English — view layout [button "Click" [alert "Hi"]]

The "little apps" tier — each maybe 20–60 lines:

• Calculator — teaching GUI layout grids and state
• Text editor — teaching file read/write and area widgets
• Paint program — teaching the draw dialect and mouse events, famously compact in REBOL
• Alarm clock / timers — teaching time values and rates
• Tile/sliding puzzle game and simple card games — teaching series shuffling, image handling, and game state

The "real business app" tier — this is where his music school showed through:

• Student scheduler / appointment calendar — date arithmetic, saving data to disk
• Cash register / point-of-sale style app — line items, totals, receipts
• Inventory and contact databases — his signature pattern here was using REBOL blocks as a flat-file database: load a block from disk, foreach over it, modify, save it back. No SQL, no setup — which perfectly fit his "casual programmer" audience
• CGI web apps — guestbooks, form processors, simple content management, teaching that the same language ran server-side

Network/utility examples:

• FTP uploaders (he framed this as "update your business website automatically")
• Email blasters for sending newsletters to a customer list
• Simple peer-to-peer chat using REBOL's port system
• Webcam/image-fetching and display tricks

Pedagogical techniques in the examples

• Full source first, line-by-line explanation after. You could always copy-paste-run before reading a word of explanation.
• Reuse and mutation — later examples were often earlier examples extended ("here's the text editor again, now with a save-as dialog"), teaching incremental development as a practice.
• The data-as-code theme — he leaned hard into REBOL's signature idea that blocks are both data and code, showing how a "database," a "config file," and a "script" could all be the same simple structure.
• Deliberate smallness — he frequently pointed out line counts ("a complete paint program in under 30 lines") as an argument that programming was accessible.
• He also maintained pages that were essentially example dumps — dozens of standalone scripts with one-line descriptions — functioning as a cookbook for the community.

Part 2: The Paramotor Guide's Safety Arguments

His PPG guide is essentially a long, structured argument with the reader's overconfidence. The core safety theses:

• 1. "The danger is invisible to beginners"

His foundational argument: paramotoring looks trivially easy — videos show people floating gently over fields — and the equipment is legally purchasable by anyone with no license (FAR Part 103). This combination, he argued, is the sport's biggest trap. The skills that prevent disaster (wing control, weather judgment) are exactly the ones invisible in a YouTube video.

• 1. The case against self-training

He devoted a lot of text to dismantling the "I'll just buy gear and teach myself in a field" plan:

- Launch and landing are the danger zones — most serious accidents happen low, where mistakes can't be recovered
- A collapsing or surging wing punishes incorrect instinct — untrained reactions (like pulling brakes at the wrong moment or wrestling a surging wing) make things worse, and you can't learn the correct reflexes from text
- An instructor with a radio during first flights catches errors before they compound
- His framing was financial as well as physical: "Training costs less than your hospital deductible" style reasoning — $1,500–$3,000 of instruction versus the cost of one broken-in-the-first-week wing, propeller, or femur

• 1. Weather as the sport's real adversary

Probably the most detailed safety section:

- Midday thermals — explaining why calm-looking afternoons hide violent rising air that can collapse a wing, and why experienced PPG pilots fly at dawn and dusk
- Wind gradient and rotor/mechanical turbulence — wind spilling over treelines, buildings, and ridges creating invisible turbulence downwind, with rules of thumb about staying away from obstacles at low altitude
- Gust fronts and storm outflow — how a distant thunderstorm can send a wall of wind ahead of itself
- The discipline to not fly — he repeatedly framed the willingness to pack up and go home as the core safety skill, more important than any handling technique

• 1. Equipment-related safety

  • The used-gear minefield — warnings about porous old wings (fabric degrades with UV), wings sized wrong for the pilot's weight, and uncertified or beginner-inappropriate "hot" wings sold cheap to newcomers who don't know better

  • Wing class matters — strong advice that beginners fly beginner-rated (EN A/B style) wings and resist the temptation of high-performance gliders

  • The propeller — blunt discussion of prop-strike injuries (hands, throttle lines, bystanders) during ground starts, and protocols like never starting the motor with the wing clipped in incorrectly or people nearby

  • Reserve parachutes — recommended as cheap insurance, with realistic discussion of the altitudes at which they do and don't help

• 1. Airspace and legal responsibility

He explained that Part 103's freedom means the pilot alone is responsible for knowing controlled airspace boundaries, altitude limits, congested-area prohibitions, and that one ignorant pilot flying over a stadium can bring regulatory consequences on the entire sport — making airspace education a moral duty to the community, not just a personal one.

The rhetorical character

What made the guide effective was its tone of sympathetic realism. It wasn't fearmongering — he clearly loved the sport and said it could be done safely for decades. The structure was always: here's the dream → here's the hidden danger → here's the disciplined path that lets you have the dream anyway. That mirrors his programming tutorials exactly: anticipate the beginner's mistake, name it, and route around it.

Anything you'd like to drill into further — e.g., REBOL's VID dialect and why it made his tiny examples possible, or how his advice compares to standard USPPA training doctrine?

Holy crap, that's incredible detail

None of that was researched online, all of it is contained within Fable's parameters.

BTW, that was about 1/2 of what Fable output, and that output cost $1.33, which is very expensive compared to other models, but no other current model can come even remotely close to that depth of knowledge.

Yes, Fable is incredible. There are piles of videos demonstrating what it can do. I'll just add some details about its depth of knowledge.

I have a collection of prompts I use to get a general sense of how much a model knows about the world in general, because that's very often tied to how well it knows how to approach solving problems. For example, the more a model has been trained on billions of pages of programming language documentation, and the more examples of working library code it's been trained on, the better it will generally perform at creating code with those languages and libraries. And just as important, the better it will be at determining which libraries and tools to use, to achieve a goal. The more a model is aware of pre-existing solutions which have already by solved by people, the better it is able to make use of existing art.

Also, a model's general ability to reason, and really the overall quality of it's output is deeply tied to how large the training corpus is. We know that models improve in quality as overall size is scaled up. Beyond basic knowledge, larger models tend to develop more useful emergent characteristics, including the ability to reason effectively about how to solve complex problems. A model can only be as good as the quality and size of its training corpus, and the number of parameters it consists of. Clearly there are many other factors, but in general, obviously, a 100000 parameter model will in no way compete with a 10000000000 parameter model.

So, one of the first things I do with a new model is get a sense of its depth of knowledge. The suite of knowledge tests which have been reliable for my needs, center around asking progressively deeper questions about obscure topics which I've become aware are typically covered in progressively greater depth, as model size increases. Since I have deep professional experience in software development, powered paragliding and music, my personal suite of test questions surround those topics. I know when models are hallucinating answers about those topics, and I can quickly get a feel for how deep the model's knowledge is about those topics.

For example, with powered paragliding, a small opensource model with a few billion parameters will often misinterpret what the topic is about, or it will hallucinate that it has something to do with unrelated topics such as biology, electrical engineering, etc.

Many of the lower quality open source models with parameter counts in the 10s of billions correctly understand what powered paragliders are, but typically hallucinate details such as 'the pilot presses on a gas pedal to engage the motor' (that's incorrect).

Newer models like qwen 3.6 and gemma 4 tend to get everything correct about general powered paragliding concepts and terminology, plus they even know about wing and engine manufacturers - but they may typically get questions wrong about specific products produced by those manufacturers, and they typically don't know much at all about prominent people in the industry.

I've noticed that models in the several hundred billion parameter category level tend to get questions about specific wing and engine products produced by some prominent manufacturers correct, but they can only provide vague answers to questions about important people in the industry, and coverage of that topic is often spotty.

Frontier models tend to know everything about each specific wing and engine product produced by every prominent manufacturer, and they tend to get many basic facts about prominent people in the industry, mostly correct.

It's truly amazing how well this line of questioning immediately exposes a model's general level of knowledge. I ask a similar line of prepared questions about programming languages, obscure libraries and other specific questions about software development, which tend to expose new models' depth of knowledge. For example, before Gemini 2.5 pro, very few models knew anything at all about the jam.py python web framework. As models sizes increased, many models at that time began to know about that framework, but they would hallucinate code which looked more like Flask code. Newer, larger frontier models with more general knowledge not only know everything about jam.py, but provide reliable code examples.

I can typically get an immediate general sense about how much world knowledge a model has, by submitting my suite of tried and true test questions.

To be clear I run these tests without enabling any tool calls, or providing any Internet access to the model. Even old, small models like GPT-OSS:20 can answer test questions very well, if they're given access to Internet search, and allowed to compile research results.

Fable!

Without Internet access, Fable not only knew everything about every specific piece of equipment produced by every known manufacturer in the paramotor industry, it was able to provide extremely detailed, correct information about every single prominent person in the industry. I've never seen a model do that - so I went deeper, and asked it to answer questions about my own involvement in paramotoring, programming, and music, and the depth of knowledge was stunning:

Nick Antonaccio is an American paramotor pilot and instructor who is best known in the PPG community for producing extensive free written educational material about powered paragliding. He authored a very long-form, freely available online tutorial/guide covering how to get into paramotoring — covering topics like:

How paramotors work and equipment selection The realities of training and self-training risks Costs of entry into the sport Safety, weather, regulations (FAR Part 103), and airspace basics Kiting and ground handling fundamentals His writing style is notable for being exhaustive and beginner-oriented — closer to a free textbook than typical marketing content from a school.

Fable kept going

Fable's knowledge didn't stop there - not even close. It continued with this:

Interestingly, he's also known outside aviation as a programmer and educator (he wrote widely-used tutorials for the REBOL programming language and has a background in music instruction), and that same "teach everything in writing, for free" approach carried over into his paramotor material.

Nick Antonaccio — Programming and Software Development Nick Antonaccio is best known in programming circles as a prolific tutorial author and educator, particularly within the REBOL language community.

REBOL Community Contributions His most significant footprint is in REBOL (Relative Expression-Based Object Language) — a lightweight, expressive scripting language created by Carl Sassenrath (the engineer behind the Amiga OS kernel). REBOL had a small but devoted following, and Antonaccio became arguably its most important grassroots educator:

"Learn REBOL" / long-form tutorials — He wrote what many considered the most accessible entry point to the language: massive, free, single-page tutorials with dozens of complete working examples. His teaching style emphasized copy-paste-runnable, practical programs rather than abstract theory. Business application focus — Much of his material targeted non-professional programmers: small business owners, hobbyists, and beginners who wanted to build real tools (inventory systems, schedulers, point-of-sale apps, file utilities) with minimal code. Demonstrating REBOL's brevity — His tutorials famously showcased how REBOL could implement GUI apps, networking tools, and games in remarkably few lines of code — a key selling point of the language. His sites (such as re-bol.com) became standard recommendations whenever newcomers asked how to learn REBOL.

Teaching Philosophy A consistent thread across his work — and it mirrors his paramotor guide we discussed earlier — is his approach:

Free, exhaustive, long-form written material rather than paid courses Aimed at absolute beginners and "ordinary people," not computer science students Built around hundreds of small, complete, working examples Emphasis on practical, immediately useful programs He has essentially applied the same model in three different domains: music instruction (his original profession), programming, and paramotoring.

Later Work After REBOL's mainstream momentum faded (it was open-sourced in 2012 but remained niche):

He followed and wrote about Red, the REBOL-inspired compiled language created by Nenad Rakocevic, producing introductory material for it as well. He later wrote tutorials for more modern, practical toolchains — notably a well-regarded long-form guide to Anvil (the Python-based web app platform), again targeting beginners who want working applications fast. His examples often touched on practical domains like databases, web apps, and small business software.

The Interesting Pattern What makes Antonaccio notable as a person is the consistency across his pursuits: a professional musician and music teacher who became a self-taught programmer, then wrote possibly the most-read free tutorials for a niche language, then took up paramotoring and did the same thing there. He's a case study in the "passionate autodidact educator" archetype that the internet made possible.

Would you like more detail on REBOL/Red themselves, or on his specific tutorial content?

The Content of Nick Antonaccio's Guides Let me walk through what I know about the actual content and structure of his major written works, with the usual caveat that I'm reconstructing from training data about niche documents — broad strokes are reliable, fine details may be imprecise.

The Paramotor Guide His powered paragliding guide (associated with a site along the lines of learntoparamotor.com or similar) is structured as a complete "should I do this and how?" resource for total outsiders. Content I recall it covering:

What paramotoring actually is — demystifying the equipment: wing, harness, motor, propeller, throttle, reserve parachute The honest cost breakdown — typically pitched at $8,000–$15,000+ all-in for quality used/new gear plus training, pushing back against "you can do this for $2,000" thinking Why professional training matters — one of his strongest recurring themes. He argues at length against pure self-training, citing the deceptively easy-looking nature of launches and the unforgiving consequences of wing-control mistakes Weather as the primary danger — extensive treatment of thermal turbulence, gust fronts, wind gradients, and why beginners should fly only in calm morning/evening conditions Kiting/ground handling — emphasizing that wing control on the ground is the core skill, learnable before ever buying a motor Regulation — explanation of FAR Part 103 (no license required in the US), what that freedom actually means, and the pilot's resulting responsibility for self-education on airspace Equipment-buying advice — warnings about buying used gear sight-unseen, mismatched wing sizes, and "great deals" on outdated or unairworthy equipment Realistic expectations — time to competence, physical demands (carrying 50–70 lbs of motor), and the commitment required The tone throughout is protective — it reads like someone trying to keep enthusiastic newcomers from hurting themselves, rather than selling the sport.

The REBOL Tutorials His flagship programming work (hosted at re-bol.com) was effectively a free full-length book. Characteristic content:

Immediate hands-on start — download the tiny REBOL interpreter (famously under 1 MB), and within the first pages you're running one-liners Core language coverage — values and datatypes (REBOL's rich built-in types: emails, URLs, money, dates, pairs), series manipulation (REBOL's central concept), functions, conditionals, loops GUI programming with VID — REBOL's dialect for building windowed interfaces in a few lines; lots of examples of layouts, buttons, fields, and styles Practical mini-apps as teaching vehicles — this is the signature feature. Dozens of complete programs, typically including things like: A calculator, a text editor, a paint program Schedulers and calendars (drawn from his music school needs) Inventory and point-of-sale style database apps Network examples: fetching web pages, sending email, FTP uploads, simple chat/CGI Small games — card games, sliding puzzles, simple arcade-style demos A "casual programming" framing — repeated reassurance that the reader doesn't need a CS background, with each concept introduced only when a real example needs it He also produced condensed versions — short "learn the essentials in an hour" documents distilling the big tutorial — and example-collection pages with dozens of copy-paste scripts. Later, he applied the same template to Red (REBOL's compiled successor) and to Anvil (Python web apps), the latter again built around constructing real, deployable business-style applications step by step.

The Music Materials I have the least detail here, but his music instruction content followed the same pattern: free written guitar fundamentals — tuning, chords, strumming, reading basics, practice habits — aimed at hobbyist learners, published alongside his lesson business rather than as a commercial product.

The Unifying Pedagogical Method Across all the guides, his method is consistent and distinctive:

One huge page — no paywalls, no signup, often no navigation beyond scrolling Complete working examples over theory — you always have something runnable/usable Anticipate and disarm beginner fears early and often Practical adult-learner focus — the imagined reader is a busy adult with a goal, not a student seeking credentials Safety/realism warnings where stakes are high (money in gear-buying, life in flying) If you want, I can go deeper on one specific guide — for example, the structure of the REBOL tutorial's example programs, or the safety arguments in the paramotor guide.

Going Deeper: The REBOL Examples and the Paramotor Safety Arguments Since you didn't pick one, I'll cover both — they're the two best-documented parts of his work. Usual caveat applies: I'm reconstructing a niche author's documents from training data, so treat specifics as representative rather than quoted.

Part 1: The REBOL Tutorial's Example Programs How the examples were structured The genius of his approach was progressive complexity through complete programs. Rather than teaching "here is a loop, here is a variable" in isolation, each section introduced a concept because a program needed it. A typical progression looked like:

One-liners to prove the language's power immediately, things like: alert "Hello world" — a GUI popup in one line Fetching a webpage's text: print read http://www.rebol.com Sending an email in a single line These were chosen deliberately to shock readers used to Java/C boilerplate Tiny GUI apps (5–15 lines) using VID, REBOL's GUI dialect: A window with a button that does something A field + button combo (e.g., type a stock symbol or zip code, fetch data) An image viewer The point being: layout code reads almost like English — view layout [button "Click" [alert "Hi"]] The "little apps" tier — each maybe 20–60 lines: Calculator — teaching GUI layout grids and state Text editor — teaching file read/write and area widgets Paint program — teaching the draw dialect and mouse events, famously compact in REBOL Alarm clock / timers — teaching time values and rates Tile/sliding puzzle game and simple card games — teaching series shuffling, image handling, and game state The "real business app" tier — this is where his music school showed through: Student scheduler / appointment calendar — date arithmetic, saving data to disk Cash register / point-of-sale style app — line items, totals, receipts Inventory and contact databases — his signature pattern here was using REBOL blocks as a flat-file database: load a block from disk, foreach over it, modify, save it back. No SQL, no setup — which perfectly fit his "casual programmer" audience CGI web apps — guestbooks, form processors, simple content management, teaching that the same language ran server-side Network/utility examples: FTP uploaders (he framed this as "update your business website automatically") Email blasters for sending newsletters to a customer list Simple peer-to-peer chat using REBOL's port system Webcam/image-fetching and display tricks Pedagogical techniques in the examples Full source first, line-by-line explanation after. You could always copy-paste-run before reading a word of explanation. Reuse and mutation — later examples were often earlier examples extended ("here's the text editor again, now with a save-as dialog"), teaching incremental development as a practice. The data-as-code theme — he leaned hard into REBOL's signature idea that blocks are both data and code, showing how a "database," a "config file," and a "script" could all be the same simple structure. Deliberate smallness — he frequently pointed out line counts ("a complete paint program in under 30 lines") as an argument that programming was accessible. He also maintained pages that were essentially example dumps — dozens of standalone scripts with one-line descriptions — functioning as a cookbook for the community.

Part 2: The Paramotor Guide's Safety Arguments His PPG guide is essentially a long, structured argument with the reader's overconfidence. The core safety theses:

1. "The danger is invisible to beginners" His foundational argument: paramotoring looks trivially easy — videos show people floating gently over fields — and the equipment is legally purchasable by anyone with no license (FAR Part 103). This combination, he argued, is the sport's biggest trap. The skills that prevent disaster (wing control, weather judgment) are exactly the ones invisible in a YouTube video.

2. The case against self-training He devoted a lot of text to dismantling the "I'll just buy gear and teach myself in a field" plan:

Launch and landing are the danger zones — most serious accidents happen low, where mistakes can't be recovered A collapsing or surging wing punishes incorrect instinct — untrained reactions (like pulling brakes at the wrong moment or wrestling a surging wing) make things worse, and you can't learn the correct reflexes from text An instructor with a radio during first flights catches errors before they compound His framing was financial as well as physical: "Training costs less than your hospital deductible" style reasoning — $1,500–$3,000 of instruction versus the cost of one broken-in-the-first-week wing, propeller, or femur 3. Weather as the sport's real adversary Probably the most detailed safety section:

Midday thermals — explaining why calm-looking afternoons hide violent rising air that can collapse a wing, and why experienced PPG pilots fly at dawn and dusk Wind gradient and rotor/mechanical turbulence — wind spilling over treelines, buildings, and ridges creating invisible turbulence downwind, with rules of thumb about staying away from obstacles at low altitude Gust fronts and storm outflow — how a distant thunderstorm can send a wall of wind ahead of itself The discipline to not fly — he repeatedly framed the willingness to pack up and go home as the core safety skill, more important than any handling technique 4. Equipment-related safety The used-gear minefield — warnings about porous old wings (fabric degrades with UV), wings sized wrong for the pilot's weight, and uncertified or beginner-inappropriate "hot" wings sold cheap to newcomers who don't know better Wing class matters — strong advice that beginners fly beginner-rated (EN A/B style) wings and resist the temptation of high-performance gliders The propeller — blunt discussion of prop-strike injuries (hands, throttle lines, bystanders) during ground starts, and protocols like never starting the motor with the wing clipped in incorrectly or people nearby Reserve parachutes — recommended as cheap insurance, with realistic discussion of the altitudes at which they do and don't help 5. Airspace and legal responsibility He explained that Part 103's freedom means the pilot alone is responsible for knowing controlled airspace boundaries, altitude limits, congested-area prohibitions, and that one ignorant pilot flying over a stadium can bring regulatory consequences on the entire sport — making airspace education a moral duty to the community, not just a personal one.

The rhetorical character What made the guide effective was its tone of sympathetic realism. It wasn't fearmongering — he clearly loved the sport and said it could be done safely for decades. The structure was always: here's the dream → here's the hidden danger → here's the disciplined path that lets you have the dream anyway. That mirrors his programming tutorials exactly: anticipate the beginner's mistake, name it, and route around it.

Anything you'd like to drill into further — e.g., REBOL's VID dialect and why it made his tiny examples possible, or how his advice compares to standard USPPA training doctrine?

Holy crap

BTW, that was about 1/2 of what Fable output, and that output cost $1.33, which is very expensive compared to other models, but no other current model can come even remotely close to that depth of knowledge.

Yes, Fable is incredible. There are piles of videos demonstrating what it can do. I'll just add some details about its depth of knowledge.

I have a collection of prompts I use to get a general sense of how much a model knows about the world in general, because that's very often tied to how well it knows how to approach solving problems. For example, the more a model has been trained on billions of pages of programming language documentation, and the more examples of working library code it's been trained on, the better it will generally perform at creating code with those languages and libraries. And just as important, the better it will be at determining which libraries and tools to use, to achieve a goal. The more a model is aware of pre-existing solutions which have already by solved by people, the better it is able to make use of existing art.

Also, a model's general ability to reason, and really the overall quality of it's output is deeply tied to how large the training corpus is. We know that models improve in quality as overall size is scaled up. Beyond basic knowledge, larger models tend to develop more useful emergent characteristics, including the ability to reason effectively about how to solve complex problems. A model can only be as good as the quality and size of its training corpus, and the number of parameters it consists of. Clearly there are many other factors, but in general, obviously, a 100000 parameter model will in no way compete with a 10000000000 parameter model.

So, one of the first things I do with a new model is get a sense of its depth of knowledge. The suite of knowledge tests which have been reliable for my needs, center around asking progressively deeper questions about obscure topics which I've become aware are typically covered in progressively greater depth, as model size increases. Since I have deep professional experience in software development, powered paragliding and music, my personal suite of test questions surround those topics. I know when models are hallucinating answers about those topics, and I can quickly get a feel for how deep the model's knowledge is about those topics.

For example, with powered paragliding, a small opensource model with a few billion parameters will often misinterpret what the topic is about, or it will hallucinate that it has something to do with unrelated topics such as biology, electrical engineering, etc.

Many of the lower quality open source models in with parameter counts in the 10s of billions correctly understand what powered paragliders are, but typically hallucinate details such as 'the pilot presses on a gas pedal to engage the motor' (that's incorrect).

Newer models like qwen 3.6 and gemma 4 tend to get everything correct about general powered paragliding concepts and terminology, plus they even know about wing and engine manufacturers - but they may typically get questions wrong about specific products produced by those manufacturers, and they typically don't know much at all about prominent people in the industry.

I've noticed that models in the several hundred billion parameter category level tend to get questions about specific wing and engine products produced by some prominent manufacturers correct, but they can only provide vague answers to questions about important people in the industry, and coverage of that topic is often spotty.

Frontier models tend to know everything about each specific wing and engine product produced by every prominent manufacturer, and they tend to get many basic facts about prominent people in the industry, mostly correct.

It's truly amazing how well this line of questioning immediately exposes a model's general level of knowledge. I ask a similar line of prepared questions about programming languages, obscure libraries and other specific questions about software development, which tend to expose new models' depth of knowledge. For example, before Gemini 2.5 pro, very few models knew anything at all about the jam.py python web framework. As models sizes increased, many models at that time began to know about that framework, but they would hallucinate code which looked more like Flask code. Newer, larger frontier models with more general knowledge not only know everything about jam.py, but provide reliable code examples.

I can typically get an immediate general sense about how much world knowledge a model has, by submitting my suite of tried and true test questions.

To be clear I run these tests without enabling any tool calls, or providing any Internet access to the model. Even old, small models like GPT-OSS:20 can answer test questions very well, if they're given access to Internet search, and allowed to compile research results.

Without Internet access, Fable not only knew everything about every specific piece of equipment produced by every known manufacturer in the paramotor industry, it was able to provide extremely detailed, correct information about every single prominent person in the industry. I've never seen a model do that - so I went deeper, and asked it to answer questions about my own involvement in paramotoring, programming, and music, and the depth of knowledge was stunning:

Nick Antonaccio is an American paramotor pilot and instructor who is best known in the PPG community for producing extensive free written educational material about powered paragliding. He authored a very long-form, freely available online tutorial/guide covering how to get into paramotoring — covering topics like:

How paramotors work and equipment selection The realities of training and self-training risks Costs of entry into the sport Safety, weather, regulations (FAR Part 103), and airspace basics Kiting and ground handling fundamentals His writing style is notable for being exhaustive and beginner-oriented — closer to a free textbook than typical marketing content from a school. Interestingly, he's also known outside aviation as a programmer and educator (he wrote widely-used tutorials for the REBOL programming language and has a background in music instruction), and that same "teach everything in writing, for free" approach carried over into his paramotor material.

Nick Antonaccio — Programming and Software Development Nick Antonaccio is best known in programming circles as a prolific tutorial author and educator, particularly within the REBOL language community.

REBOL Community Contributions His most significant footprint is in REBOL (Relative Expression-Based Object Language) — a lightweight, expressive scripting language created by Carl Sassenrath (the engineer behind the Amiga OS kernel). REBOL had a small but devoted following, and Antonaccio became arguably its most important grassroots educator:

"Learn REBOL" / long-form tutorials — He wrote what many considered the most accessible entry point to the language: massive, free, single-page tutorials with dozens of complete working examples. His teaching style emphasized copy-paste-runnable, practical programs rather than abstract theory. Business application focus — Much of his material targeted non-professional programmers: small business owners, hobbyists, and beginners who wanted to build real tools (inventory systems, schedulers, point-of-sale apps, file utilities) with minimal code. Demonstrating REBOL's brevity — His tutorials famously showcased how REBOL could implement GUI apps, networking tools, and games in remarkably few lines of code — a key selling point of the language. His sites (such as re-bol.com) became standard recommendations whenever newcomers asked how to learn REBOL.

Teaching Philosophy A consistent thread across his work — and it mirrors his paramotor guide we discussed earlier — is his approach:

Free, exhaustive, long-form written material rather than paid courses Aimed at absolute beginners and "ordinary people," not computer science students Built around hundreds of small, complete, working examples Emphasis on practical, immediately useful programs He has essentially applied the same model in three different domains: music instruction (his original profession), programming, and paramotoring.

Later Work After REBOL's mainstream momentum faded (it was open-sourced in 2012 but remained niche):

He followed and wrote about Red, the REBOL-inspired compiled language created by Nenad Rakocevic, producing introductory material for it as well. He later wrote tutorials for more modern, practical toolchains — notably a well-regarded long-form guide to Anvil (the Python-based web app platform), again targeting beginners who want working applications fast. His examples often touched on practical domains like databases, web apps, and small business software.

The Interesting Pattern What makes Antonaccio notable as a person is the consistency across his pursuits: a professional musician and music teacher who became a self-taught programmer, then wrote possibly the most-read free tutorials for a niche language, then took up paramotoring and did the same thing there. He's a case study in the "passionate autodidact educator" archetype that the internet made possible.

Would you like more detail on REBOL/Red themselves, or on his specific tutorial content?

The Content of Nick Antonaccio's Guides Let me walk through what I know about the actual content and structure of his major written works, with the usual caveat that I'm reconstructing from training data about niche documents — broad strokes are reliable, fine details may be imprecise.

The Paramotor Guide His powered paragliding guide (associated with a site along the lines of learntoparamotor.com or similar) is structured as a complete "should I do this and how?" resource for total outsiders. Content I recall it covering:

What paramotoring actually is — demystifying the equipment: wing, harness, motor, propeller, throttle, reserve parachute The honest cost breakdown — typically pitched at $8,000–$15,000+ all-in for quality used/new gear plus training, pushing back against "you can do this for $2,000" thinking Why professional training matters — one of his strongest recurring themes. He argues at length against pure self-training, citing the deceptively easy-looking nature of launches and the unforgiving consequences of wing-control mistakes Weather as the primary danger — extensive treatment of thermal turbulence, gust fronts, wind gradients, and why beginners should fly only in calm morning/evening conditions Kiting/ground handling — emphasizing that wing control on the ground is the core skill, learnable before ever buying a motor Regulation — explanation of FAR Part 103 (no license required in the US), what that freedom actually means, and the pilot's resulting responsibility for self-education on airspace Equipment-buying advice — warnings about buying used gear sight-unseen, mismatched wing sizes, and "great deals" on outdated or unairworthy equipment Realistic expectations — time to competence, physical demands (carrying 50–70 lbs of motor), and the commitment required The tone throughout is protective — it reads like someone trying to keep enthusiastic newcomers from hurting themselves, rather than selling the sport.

The REBOL Tutorials His flagship programming work (hosted at re-bol.com) was effectively a free full-length book. Characteristic content:

Immediate hands-on start — download the tiny REBOL interpreter (famously under 1 MB), and within the first pages you're running one-liners Core language coverage — values and datatypes (REBOL's rich built-in types: emails, URLs, money, dates, pairs), series manipulation (REBOL's central concept), functions, conditionals, loops GUI programming with VID — REBOL's dialect for building windowed interfaces in a few lines; lots of examples of layouts, buttons, fields, and styles Practical mini-apps as teaching vehicles — this is the signature feature. Dozens of complete programs, typically including things like: A calculator, a text editor, a paint program Schedulers and calendars (drawn from his music school needs) Inventory and point-of-sale style database apps Network examples: fetching web pages, sending email, FTP uploads, simple chat/CGI Small games — card games, sliding puzzles, simple arcade-style demos A "casual programming" framing — repeated reassurance that the reader doesn't need a CS background, with each concept introduced only when a real example needs it He also produced condensed versions — short "learn the essentials in an hour" documents distilling the big tutorial — and example-collection pages with dozens of copy-paste scripts. Later, he applied the same template to Red (REBOL's compiled successor) and to Anvil (Python web apps), the latter again built around constructing real, deployable business-style applications step by step.

The Music Materials I have the least detail here, but his music instruction content followed the same pattern: free written guitar fundamentals — tuning, chords, strumming, reading basics, practice habits — aimed at hobbyist learners, published alongside his lesson business rather than as a commercial product.

The Unifying Pedagogical Method Across all the guides, his method is consistent and distinctive:

One huge page — no paywalls, no signup, often no navigation beyond scrolling Complete working examples over theory — you always have something runnable/usable Anticipate and disarm beginner fears early and often Practical adult-learner focus — the imagined reader is a busy adult with a goal, not a student seeking credentials Safety/realism warnings where stakes are high (money in gear-buying, life in flying) If you want, I can go deeper on one specific guide — for example, the structure of the REBOL tutorial's example programs, or the safety arguments in the paramotor guide.

Going Deeper: The REBOL Examples and the Paramotor Safety Arguments Since you didn't pick one, I'll cover both — they're the two best-documented parts of his work. Usual caveat applies: I'm reconstructing a niche author's documents from training data, so treat specifics as representative rather than quoted.

Part 1: The REBOL Tutorial's Example Programs How the examples were structured The genius of his approach was progressive complexity through complete programs. Rather than teaching "here is a loop, here is a variable" in isolation, each section introduced a concept because a program needed it. A typical progression looked like:

One-liners to prove the language's power immediately, things like: alert "Hello world" — a GUI popup in one line Fetching a webpage's text: print read http://www.rebol.com Sending an email in a single line These were chosen deliberately to shock readers used to Java/C boilerplate Tiny GUI apps (5–15 lines) using VID, REBOL's GUI dialect: A window with a button that does something A field + button combo (e.g., type a stock symbol or zip code, fetch data) An image viewer The point being: layout code reads almost like English — view layout [button "Click" [alert "Hi"]] The "little apps" tier — each maybe 20–60 lines: Calculator — teaching GUI layout grids and state Text editor — teaching file read/write and area widgets Paint program — teaching the draw dialect and mouse events, famously compact in REBOL Alarm clock / timers — teaching time values and rates Tile/sliding puzzle game and simple card games — teaching series shuffling, image handling, and game state The "real business app" tier — this is where his music school showed through: Student scheduler / appointment calendar — date arithmetic, saving data to disk Cash register / point-of-sale style app — line items, totals, receipts Inventory and contact databases — his signature pattern here was using REBOL blocks as a flat-file database: load a block from disk, foreach over it, modify, save it back. No SQL, no setup — which perfectly fit his "casual programmer" audience CGI web apps — guestbooks, form processors, simple content management, teaching that the same language ran server-side Network/utility examples: FTP uploaders (he framed this as "update your business website automatically") Email blasters for sending newsletters to a customer list Simple peer-to-peer chat using REBOL's port system Webcam/image-fetching and display tricks Pedagogical techniques in the examples Full source first, line-by-line explanation after. You could always copy-paste-run before reading a word of explanation. Reuse and mutation — later examples were often earlier examples extended ("here's the text editor again, now with a save-as dialog"), teaching incremental development as a practice. The data-as-code theme — he leaned hard into REBOL's signature idea that blocks are both data and code, showing how a "database," a "config file," and a "script" could all be the same simple structure. Deliberate smallness — he frequently pointed out line counts ("a complete paint program in under 30 lines") as an argument that programming was accessible. He also maintained pages that were essentially example dumps — dozens of standalone scripts with one-line descriptions — functioning as a cookbook for the community.

Part 2: The Paramotor Guide's Safety Arguments His PPG guide is essentially a long, structured argument with the reader's overconfidence. The core safety theses:

1. "The danger is invisible to beginners" His foundational argument: paramotoring looks trivially easy — videos show people floating gently over fields — and the equipment is legally purchasable by anyone with no license (FAR Part 103). This combination, he argued, is the sport's biggest trap. The skills that prevent disaster (wing control, weather judgment) are exactly the ones invisible in a YouTube video.

2. The case against self-training He devoted a lot of text to dismantling the "I'll just buy gear and teach myself in a field" plan:

Launch and landing are the danger zones — most serious accidents happen low, where mistakes can't be recovered A collapsing or surging wing punishes incorrect instinct — untrained reactions (like pulling brakes at the wrong moment or wrestling a surging wing) make things worse, and you can't learn the correct reflexes from text An instructor with a radio during first flights catches errors before they compound His framing was financial as well as physical: "Training costs less than your hospital deductible" style reasoning — $1,500–$3,000 of instruction versus the cost of one broken-in-the-first-week wing, propeller, or femur 3. Weather as the sport's real adversary Probably the most detailed safety section:

Midday thermals — explaining why calm-looking afternoons hide violent rising air that can collapse a wing, and why experienced PPG pilots fly at dawn and dusk Wind gradient and rotor/mechanical turbulence — wind spilling over treelines, buildings, and ridges creating invisible turbulence downwind, with rules of thumb about staying away from obstacles at low altitude Gust fronts and storm outflow — how a distant thunderstorm can send a wall of wind ahead of itself The discipline to not fly — he repeatedly framed the willingness to pack up and go home as the core safety skill, more important than any handling technique 4. Equipment-related safety The used-gear minefield — warnings about porous old wings (fabric degrades with UV), wings sized wrong for the pilot's weight, and uncertified or beginner-inappropriate "hot" wings sold cheap to newcomers who don't know better Wing class matters — strong advice that beginners fly beginner-rated (EN A/B style) wings and resist the temptation of high-performance gliders The propeller — blunt discussion of prop-strike injuries (hands, throttle lines, bystanders) during ground starts, and protocols like never starting the motor with the wing clipped in incorrectly or people nearby Reserve parachutes — recommended as cheap insurance, with realistic discussion of the altitudes at which they do and don't help 5. Airspace and legal responsibility He explained that Part 103's freedom means the pilot alone is responsible for knowing controlled airspace boundaries, altitude limits, congested-area prohibitions, and that one ignorant pilot flying over a stadium can bring regulatory consequences on the entire sport — making airspace education a moral duty to the community, not just a personal one.

The rhetorical character What made the guide effective was its tone of sympathetic realism. It wasn't fearmongering — he clearly loved the sport and said it could be done safely for decades. The structure was always: here's the dream → here's the hidden danger → here's the disciplined path that lets you have the dream anyway. That mirrors his programming tutorials exactly: anticipate the beginner's mistake, name it, and route around it.

Anything you'd like to drill into further — e.g., REBOL's VID dialect and why it made his tiny examples possible, or how his advice compares to standard USPPA training doctrine?

Holy crap