Request For Hyperfixation

By: Spencer Stingley

This is a running list of & my thoughts on upcoming blog posts. I dont' know if I'll get to any of these, but I need a way to organize them. If the hyperfixation strikes, they'll become full posts. Until then, they'll live here and I'll slowly chip away on my notes.

Feel free to write about any of this! I only ask that you message me if you do. I'd love you read your thoughts on my random musings! (Reach out to me at 'contact@blankcanvas.io').

Reducing AI Costs

  • Created: 3/6/2026

Rank: 4

The price per generation going down on frontier models is helpful, as a series of apps that wouldn't have been created can be. The principle example I'm thinking of is random apps for a small business. I view AI as doing to apps what excel did for spread sheets. As time goes on, we should be able to optimize costs, making building these apps more cost effective than not, which is what happens currently. Many businesses could use a lot more software, but developer cost is prohibitive.

But as models get more effective, AI companies will want to utilize their clusters to run more powerful models, because enterprise customers pay more. This means smaller models will be competing with larger models for hardware time. This will greatly increase the cost of smaller models, simply as a product of utilizing GPU time. So the smaller models may increase in price such that the use cases for AI we are currently trying to build will be made cost prohibitive.

Just something I'm noodling around. IDK how to even begin to think about something like this. If this is right, the correct AI strategy is much longer term than GPT wrappers. Technically, you should start building things for AI models that will exist, since they would price out anything you can build with the current generation. The issue with that is its a massive risk. Fun fun fun.

Make a virtual machie

  • Created: 3/9/2026

Rank: 4

Would be a useful way to learn C++ for performance. We can make that bitch go vroom and eventually move into real virtualization techniques.

Will use 8085 assembler, since its simple. Interrupts will be weird, but whatever.

8085 reference manual: https://bitsavers.trailing-edge.com/components/intel/MCS80/9800301D_8080_8085_Assembly_Language_Programming_Manual_May81.pdf

Lets write a front end / debugger for it so people can learn the basics of assembly, computer arch, and OSes with it!

Will need to figure out a nice way to do the graphics, but whatever

It would be cool to add pipelining, branch prediction, etc. Should look into that

  • If these were optional toggles, we could teach a whole bunch about it

We could go so far as to build example programming languages for it. From language down to assembly with a visual debugger would be cool. Lots of work though...

Lets also use some CLI AI tools, so remove some of the shit work from my plate. Would be a good stress test

Plan 9 Kernel Deep dive

  • Created: 3/5/2026

Rank: 3

Look into the plan 9 kernel. The google write up on golang has the following passage, which is EXTREMELY interesting:

"One extreme example is the Plan 9 kernel, in which all system data items implemented exactly the same interface, a file system API defined by 14 methods. This uniformity permitted a level of object composition seldom achieved in other systems, even today. Examples abound. Here's one: A system could import (in Plan 9 terminology) a TCP stack to a computer that didn't have TCP or even Ethernet, and over that network connect to a machine with a different CPU architecture, import its /proc tree, and run a local debugger to do breakpoint debugging of the remote process. This sort of operation was workaday on Plan 9, nothing special at all. The ability to do such things fell out of the design; it required no special arrangement (and was all done in plain C)."

Articles to Read

  • Created: 3/5/2026

Rank: 2

Technical write ups on Google's tooling:

QoL Website Updates

  • Created: 3/5/2026

Rank: 1

Update RFH on mobile

Make individual RFH linkable, without linking the entire page. It would be nice to ask people about specific ideas and not sharing everything (since this is a slop dumping ground)

The Most TechBro Thing I've Thought

  • Created: 4/9/2026

Rank: 1

The AI arms race is an inverse bitcoin problem

Kubectl Builder

  • Created: 4/28/2026

Rank: 1

https://book.kubebuilder.io/architecture.html

Making sure this works in prod

  • Created: 2/27/2026

Rank: 0

Get absolutely baited ;)