Ghost in the Cell |
|
あ、こんにちは,
I grew up with a small, unspoken contract. It lived somewhere behind the ribs. To do well. To go farther. To justify a lifetime spent bending circumstances so mine would not have to. I did not expect that the leaving would come precisely when standing became possible, when the ground finally steadied beneath my feet. There are fractures that don’t announce themselves as breaks. They sound instead like timing. I was sent out into the world early — across continents, across climates, across borrowed languages — taught to look closely, to read everything, to collect wonder carefully, as if it might run out. I did all of that. I saw enough for two lives. But the one journey that mattered was postponed indefinitely, and now the memories carry a quiet weight they didn’t have before. It feels, sometimes, as though she waited. Long enough for the fall not to kill me. Long enough to be sure. And then stepped aside. There was a final gift — a machine built for speed, precise and unforgiving — placed gently into my hands, as if to say some distances are not meant to be walked. A friend suggested grief counselling. But grief is not an administrative problem for me. Words can't renegotiate memory, can't sand down the edges of absence. Counselling is stillness. Acceptance. The art of living smaller around what cannot be replaced. But I am not interested in becoming manageable. I am not looking for peace. I am looking for what slipped through time — or failing that, its outline, its afterimage. And there is too much ground to cover...
So now I run... I look for what I lost, searching a quarter mile at a time. Nothing else matters. Those 8 seconds that I am running... I am free. Faster... faster... faster... till the thrill of the speed overcomes the fear of death.
What you will find here...
This place began in the dark, and from time to time it will return there — but it was never meant to stay there. What you’ll find here is a moving collection of obsessions, lessons, and fragments gathered along the way: supersport motorcycles and the long, loud love affair with unadulterated speed; stories told in four-cylinder symphonies, Linux terminals, and open-source code; wandering thoughts about minds, brains, and machines; and, sometimes, quiet lessons in street photography — because freezing time is a real superpower, and I learned early how much it matters. Most of what lives here was self-taught the hard way, alone — riding, braking, then breaking, fixing transmissions and bones, coding, rebuilding — so your mileage may vary. But when it comes to cognition and computation, I was lucky: I learned from some of the sharpest minds around, and from time to time you’ll hear echoes from the laboratories and seminars of Anne Cutler, Max Coltheart, and Jerry Fodor. Think of this space as a long road at dusk — sometimes fast, sometimes reflective, sometimes reckless — but always moving forward. There is no slow lane. There are no speed limits. So scoot back on that seat, lay flat on the tank, plant your knees and elbows on the metal body, tuck your head in under the windshield... then drop a gear, open the throttle... trust me, there is no therapy like an inline-4 engine howling in 14000rpm ecstacy!
Think of this as a slice of the old internet -- before standardization and corporate douchebaggery sucked the life out of what was really supposed to be an open playfield for creativity and information exchange. Or you could think of this as some random district of Night City. Here's a non-exhaustive list of things you can expect to encounter if you hang around long enough...loitering is highly encouraged! Please don't pee in the bushes.
- Motorcycle stories, tips, and the occassional speeding video.
- My personal marginalia on cognition, computation and the nature of minds, brains [&] machines.
- Linux... A LOT OF Linux diatribes, and the occassional reminder of why Window$ and Mac O$ suck!
- Remember: GNU is Not UNIX
- Never forget: Vi(m)... Vi(m).. Vi(m).. is the number of the beast. Use eMacs!
- You can learn eMacs in one day... everyday.
- Half-baked photography advice, as I figure things out the only way I have ever been able to. By tinkering with things.
“Computer science is no more about computers than astronomy is about telescopes.”
--- Edsger W. Dijkstra
“If someone comes along with a physical theory that describes
things that happen and things that can't possibly happen and
can't make any distinction among them, it's no contribution
to physics, understanding, theory, anything.
That's LLMs.
The high tech plagiarism works as well, or badly, for
‘languages’ that humans cannot acquire (except maybe as
puzzles) as for those they can. Therefore they are telling us
nothing about language, cognition, acquisition, anything.
Furthermore, since this is a matter of principle, irremediable,
if they improve their performance for language it only reveals
more clearly their fundamental and irremediable flaws, since by
the same token they will improve their performance for
impossible systems.”
--- Noam Chomsky on LLMs
| This | is | a | table. |
|---|---|---|---|
| You | can | put | columns |
| and | rows | in | tables. |
| Tables | are | very | SEXY! |
Why, you ask?
Noam Chomsky on Guiding Principles
Noam Chomsky on the disconnect of the Left with Reality
Noam Chomsky on Woke Nonsense and Identity Politics
Noam Chomsky on Life, Information, Language and Intelligence
“Large parts of the Left have abandoned rational inquiry.”
— Noam Chomsky, interview with Lingua Franca, 1992
“The Right has long been deeply hostile to science.”
— Noam Chomsky, interview on science and politics
“Relativism is a way of avoiding the use of reason.”
— Noam Chomsky, public lecture
What I used to ride
Suzuki GSX-R 1000R
- Turbocharged, Inline-4 Cylinder 999CC Engine.
- 195 horsepowers at the rear wheel.
- 130 Nm of torque @ 13,000 RPM.
- Top speed: ~327 km/h.
For a short while, in Montreal, I had access to the legendary 2005 Suzuki GSX-R 1000R. And this one was TURBOCHARGED!! The fastest I ever went was around 264 km/h. But it was capable of so much more… sigh
What I currently ride
Aprilia RS 457
- 275 Degrees Cranked Inline-2 Cylinders Engine.
- Custom ECU tune by MC Customs.
- 62 horsepowers at the rear wheel.
- 58 Nm of torque @ 10,500 RPM.
- Top speed: ~206 km/h.
A step down from what I used to ride overseas—but not at all bad. The inline-4 GSX-R delivers explosive top-end power, while the 275° parallel-twin RS oozes mid-range torque. On Indian roads, you’ll never stretch the GSX-R past second or third gear—but you can use all of the RS 457’s power. And that’s where the fun is. You know what they say... it's a lot more fun pushing a relatively slower bike to its extreme than it is to ride a blisteringly fast bike at mid capacity. A gift from my Ma...
What I want to ride next
Suzuki GSX-R 1340RR, Hayabusa
- Bow to The Queen!
- 1340CC Inline-4 Cylinders Hayabusa Engine.
- 215 Horsepowers at the Rear Wheel.
- 158 Nm of torque @ 14000 RPM.
- Top speed: 370 km/h (Digitally Unrestricted)...
What can I say? She is the undisputed Queen of motorcycles! The fastest production motorcycle in the world since 1996. Three decades at the top of the food chain!
What computes
From Freedom, Came Elegance
- Ryzen 9 9900X.
- Zotac Amp Extreme Infinity RTX 5080.
- 2x16 GB Corsair Dominator DDR5 @ 6600Mhz.
- Aorus Riptide Wifi X870
- Running on Ubuntu 25.10 Gnome Edition
Computing is more than just using a computer. It is knowledge and information, in the rawest sense. And knowledge should be Free (libre and gratis!!). Like many of us in this country Window$ was forced upon me, and the carrot of Mac O$ was dangled in front. I was supposed to use the former and aspire to the latter. Then, sometime in the mind-2000s, thanks to a wonderful magazine called Digit, I stumbled upon an operating system called GNU/Linux. It came in many flavors, called distributions (because they were freely distributed!), and the first one that I chanced upon was a (now extinct) Japanese distribution called Turbolinux. Initially the attraction was just that it was different, different from the same shit that everyone around me was using. It bothered me... still does... that so many people never ask if things could be another way? In 2006 I discovered that there was (still is!) this company called Canonical... and that they made this Linux distribution called Ubuntu(an African concept meaning Humanity towards others.)... and that they even had a program called Ship It where you could visit the website, fill in your address and Canonical will ship you a CD of the latest Ubuntu to your address. Anywhere in the world. FOR FREE! Over years I acquired three such discs from Canonical (see picture above!), and it got me started down a path of free and open source software that I still walk. I have never looked back. Though economic reasons have long since forced Canonical to abandon the Ship It program, it's legacy lives on. It taught so many of us, by example, that software should not be behind paywall... that code, like recipes, should shared, not hoarded. People over profit, as Noam Chomsky would say. In these days of instant download and fast internet, Ship It seems almost archaic! But I can't lie... there was something about being able to unpack those envelops from Canonical, trying to anticipate what they new case design would look like!
What computes
From Structure, Came Meaning
- Language as a biologically grounded cognitive system.
- I-language over E-language.
- Generative grammar and constraint-based explanation.
- Acquisition as growth, not induction.
- Computation as internal organization, not behavior.
What I work on is the idea that natural language is, at its core, a computational system with mathematically characterizable properties. Human languages are not arbitrary symbol streams shaped by usage alone; they are constrained generative systems that define precise sets of possible structures and systematically exclude vast swathes of imaginable ones. These constraints are not learned by induction over data. They reflect an internal architecture that sharply limits the hypothesis space before experience ever enters the picture.
From this perspective, syntax is not an emergent epiphenomenon of communication, but a formal object: a rule-governed system capable of unbounded hierarchical composition from finite means. The central questions are therefore computational ones. What kinds of operations are permitted? What representations are available? What class of functions does the system compute? And why do human languages cluster so tightly within a narrow region of this abstract design space?
These questions cannot be answered at the level of surface behavior or corpus statistics. They force engagement with the neurobiology of language itself. Work on cortical circuitry, timing constraints, and cross-species comparison suggests that syntactic computation is implemented by specific neural mechanisms rather than diffuse associative processes. The relevance of birdsong here is not metaphorical: songbirds exhibit structured vocal sequences governed by internal grammars, offering a biologically tractable window into how sequence-generating systems are realized in neural tissue.
Comparative work on birdsong highlights both continuity and discontinuity. While birds lack the full combinatorial power of human syntax, they share key properties: discrete symbolic elements, rule-governed sequencing, and developmental constraints that cannot be reduced to imitation alone. These parallels matter because they suggest that the building blocks of syntactic computation are biologically grounded, not culturally invented, and that language evolution is constrained by what neural systems can implement.
This biological grounding has immediate consequences for artificial intelligence. Large language models excel at surface imitation, but they lack internal symbolic structure, explicit variables, and stable representations of hierarchical dependency. As a result, they blur the distinction between what is statistically likely and what is structurally possible. They generate fluent outputs, but they do not possess grammars. Performance improves, but explanation does not.
Neurosymbolic AI is therefore not a nostalgic return to hand-coded rules; it is a recognition of a basic mismatch. If human language is computed by systems that integrate symbolic structure with neural dynamics, then models that discard structure in favor of scale will inevitably hit principled limits. The failure is not one of data or compute, but of representational adequacy. Without explicit structure, there can be no genuine explanation of linguistic competence.
The broader aim of this work is to understand how abstract computation is physically realized in biological systems, and what that tells us about the nature of mind. Language is the sharpest tool we have for probing this relationship, precisely because it sits at the intersection of mathematics, biology, and cognition. Treating it seriously means refusing to confuse successful imitation with understanding, and refusing to confuse behavior with mechanism.
What I shoot with
I currently use a Fujifilm X-T5 40 MP APS-C CMOS Sensor mirrorless camera, paired with either a Fuji 16-80mm f/1:4 or a Sigma 60-600mm f/4.5-6.3 DG DN lens. Sometimes I also use my Moto Edge 60 Pro to get a few quick shots in. One thing I have found is that learning to frame a shot in your mind and visualizing the light and shadow balance that you want is more important, and a heck of a lot more difficult a skill to master, than picking a great camera. Cameras only record information... it's your mind/brain that sees.