Nano-Blog

A collection of short items related to the culture and practice of software development.

§ September 16, 2026

AI or IA?

Gibson Musical Instruments paid to have me talk to Doug Engelbart for an afternoon back in the mid-2000s. Dude was pretty old, but still sharp. He recounted arguments he and Seymour Papert had with Minsky and Pinker over the nature of consciousness and the mind. And perhaps more importantly how to use computers as "bicycles for the mind."

Engelbart landed squarely in the Don Norman, William Ross Ashby, Stafford Beer, Licklider camp. Humans will internalize the problem domain and apply tools to the models they create in order to manipulate the model or solve for a particular root (forgive the math analogy, it's late.) Minsky, et al thought we would (like LLMs today) use natural language to query and command digital systems.

I'm much more of an Intelligence Amplification (IA) fan than an Artificial Intelligence (AI) fan. Here's a reference: Intelligence Amplification [Wikipedia].

#ai #Doug-Engelbart #intelligence-amplification
About Inkling

Thinking Machine Labs published an open-weight model. This seems like a bit of a big deal to me and I sort of think this was part of their deal all along. I see guile everywhere, so I would love to think Murati slipped out the side door at OpenAI with a story about interactive AI (which OpenAI might not have been horribly concerned about.) But then when the time was right they published Inkling, their open-weight model. (The Wikipedia also has a page on the Inkling Large Language Model.)

At 600GB - 2TB, Inkling is a bit large for me to put on any of the machines I happen to have, but it's interesting to see a company in the states jump into this product (project?) space. I had always assumed Anthropic and OpenAI didn't release their model weights because a) they believed that's where their "unfair advantage" lie and b) lawyers could use it to prove they violated copyright when they trained their models.

It's interesting to note the Chinese seem to have universally adopted open-weight models, presumably because a) they were playing catch-up for a bit and wanted to seed their entire value chain with useful tools and b) there are places in the world that care more about copyright than china.

#ai #Inkling #llms #Thinking-Machines-Lab
About FemtoAI

Speaking of AI... I like to check in with Femto AI every now and again. I haven't heard anything ground-breaking from them for a while, but I love the promise: same AI goodness, less power consumption. Either makes your data-center less of a target for the torches and pitch-forks crowd or lets you run inference on mobile devices. Or both.

And they have silicon. I guess I'm a latent chip fetishist.

But I don't see Kwabena Boahen on their board any more, which is disappointing. (Dr. Boahen also has a profile page on Stanford's web site.) He's the reason I was interested in Femto to begin with.

#Kwabena-Boahen #Femto-AI
Are LLMs Conscious?

Assuming we know what we're talking about when we use words like "consciousness" and "awareness" (much less "sapience" and "sentience"), can we have a discussion about LLM consciousness without veering into the realm of science fiction? I mean, I'm a science fiction fan, so it doesn't take much to point me in that direction.

But... I was having a conversation with a dude who said he was sure LLMs were conscious. On the face of it, it's a little absurd. LLMs are statistical models which guess the best next word in a sequence. Like Markov Chains, they're word guessing models. But my comment to the guy was "The only way LLMs are conscious is if consciousness is embedded in the structure of language."

And then I stopped and thought about what I said. I meant it to be dismissive of the idea of LLM consciousness. I still don't think LLMs are conscious, but I think a type of awareness is imprinted on infants via language. Though I claim no special familiarity with case studies of feral children, a common theme arises that without early exposure to human language, "something" doesn't develop. I would call it an "awareness" rather than "consciousness," but there are few who wish to join me as I carve semantics so finely.

So my take on it is language modifies existing neuronal circuitry to create a perception of reality that predisposes one towards behaviours which are interpreted internally as consciousness. Like a virus that depends on the existing cellular infrastructure to (among other things) propagate itself, language makes use of what's already there to do things (like filter the phenomenal world, revealing a world which reinforces the ego dream) and then propagates itself via the next generation.

Language, it seems, truly is a virus.

#ai #consciousness #llms
Neuromancer Returns

Apple TV released a trailer for their upcoming production of Gibson's Neuromancer. It looks very pretty. I went back to re-read the story and was happy to revisit the sprawl. But what struck me most was how Burroughsesque it was. After a little bit of digging I discovered that wasn't accidental. In an interview with Gibson by Larry McCaffery he (Gibson) mentions Burroughs and Ballard as explicit influences.

I first read the sprawl trilogy before reading Burroughs or Ballard. Now that I'm revisiting Neuromancer, it's more obvious how much he explicitly owes (stylistically) to the past masters.

#William-Gibson #Neuromancer

§ August 24, 2025

Valve Handbook as a Carrier of Culture

Over a decade ago the Valve published a "Handbook for New Employees". It's gone on to be debunked as a handbook, but instead more of a recruiting tool. (See the discussion at Hacker News for more info.) It may not describe the culture Valve has, but certainly describes the culture they want prospective employees to think they have.

Title Page of Valve's Handbook for New Employees

Either way it's worth a read: VALVE: HANDBOOK FOR NEW EMPLOYEES. It is said that "Culture Eats Strategy for Lunch" (or Breakfast, depending on where you source the quote from.) And there's probably a lot to be gained from looking at how Valve wants prospective employees to view its culture.

#culture #software-eng

§ August 15, 2025

The Coming Software Apocalypse

Nearly a decade ago, James Somers, wrote an article for The Atlantic about how software bugs scale in relationship to system complexity. The article, The Coming Software Apocalypse, despite being 8 years old, doesn't seem that dated. Written before AI went mainstream, it focuses on how (at a high level) difficult bugs creep into software systems and proposes a few ideas luminaries in our industry are proposing.

Written for a general audience, programmers will find it lacks specific details like "where can I download these tools?" or "how do I implement these ideas in my emacs-focused workflow?" But it's still worth reading as an overview, a warning and a demonstration we're still in the software crisis.

P.S. - James also maintains a list of items he's written for mainstream publications and books he's read. It's equal parts self-promotion and a travelogue for literary odyssey. Would that more people would have the presence of mind to share such things.

#James-Somers #papers #The-Atlantic

§ August 14, 2025

Inventing on Principle

Bret Victor is known for many things. The worrydream.com website (with it's excellent list of seminal papers) and DynamicLand spring immediately to mind. But he's also known for delivering a very good presentation about a better future for software designers. The talk from 2012 is available on YouTube:

Worth a watch by anyone curious about how we may think about software and it's development.

#Bret-Victor #design #video

§ August 4, 2025

Another Small Lisp

Nary a month goes by that we do not come across another small Lisp online. And why not? Like FORTH, Lisp is a simple language to implement and many developers code their own as a way of learning the details of the language.

We previously covered two small Lisps: April 24, 2025 : A Very Small Lisp in 99 Lines of C and May 20, 2025 : On the Subject of Pico-Lisp. The Lisp we are reporting on today looks similar to van Engelen's small lisp in C, but is its own creation. If you're interested in implementing your own Lisp (and who isn't these days?) examining several previous implementations is probably not wasted effort.

The implementation we're reporting on here, Micro-Lisp is currently hosted on @sepisoad's github account, but retains A. Carl Douglas' copyright notice. I suspect this is because it's no longer available on @carld's corner of github. But the Wayback Machine archived a copy of a decent description of what's going on with the code at https://web.archive.org/web/20230922184900/https://carld.github.io/2017/06/20/lisp-in-less-than-200-lines-of-c.html

Worth a read if you're hip to writing your own small Lisp.

#lisp #programming

§ July 28, 2025

Two-Bit History Looks at Lisp

Software developer and writer Sinclair Target penned a very nice essay attempting to answer the question "Why do some programmers really like Lisp?"

Jumping between that famous XKCD comic about Lisp to a bit of history about John McCarthy to Terry Winograd, Target eventually forms a three pronged thesis. I won't give away the ending. It's probably fine to just say it's a merry jaunt through computer programming language design history and leave it at that. Worth the time it takes to read for the community interested in such things.

The text is available at Two-Bit History under the title How Lisp Became God's Own Programming Language.

#lisp #Sinclair-Target