joocifer
banner
joocifer.bsky.social
joocifer
@joocifer.bsky.social
Systems programmer specializing in cybersecurity and rabbit holes

Mass in our manifold falls toward its local center of gravity and, at the same time, toward the single point outside of our manifold where the Big Bang originated.

- Lucifer’s Topology
As long as you stick to the true underlying theme that warms our hearts even more than blouse-watch:

—> Things are never random and there is an underlying, controlling force behind everything. Even if it’s malevolent, that’s preferable to nothing, preferable to random.
November 11, 2025 at 1:16 PM
“It’s the candidate, stupid.”

This same electorate - an even more boomer-dominated electorate that we have now - voted for a black man named Barack Obama. Twice.

Very, very, very simple answer to “why”:

he was a more popular, exciting candidate than his opponents.
November 11, 2025 at 1:08 PM
Reposted by joocifer
Here’s the working model:

The wishes of the D party - which represent the greatest number of Americans, by far, when you map wants/needs - that’s DNA.

We want to inject this DNA into our future, save it in our body politic as something that shapes us.

We have to win to do that.

Here’s >
October 28, 2025 at 12:43 PM
..not yet reviewed results, bit seem to work fine, sonnet returned the result faster with less rumination -

which may be why even though GLM is 1/3 the cost of Sonnet, it it's cost was only $0.13 vs $0.20 - Sonnet seems to be more efficient, so it's less than 2x the actual cost vs 3x?
November 9, 2025 at 3:13 PM
Did a little experiment - same prompt to create a small useful tool using GLM 4.6 and Sonnet 4.5 to see cost breakdown.

First is the token usage and cost from asking GLM 4.6.

Second's the accrued usage and cost after asking Sonnet 4.5 to do the same task, so 13c for GLM, 20c for sonnet.

I've >
November 9, 2025 at 3:11 PM
For example, let’s casually throw out that the the social contract may need to change.

Minor technological step advances don’t require worrying about UBI in a public article.
November 9, 2025 at 4:31 AM
Lindsay Graham is the worst kind of bootlicker, one who realizes his wrong he is but gets down in the first to grovel anyway.
November 8, 2025 at 1:53 PM
I think it's doing tha now even without being told to make a nice commit message, it's making an almost perfect very succinct summary by default if allowed to do it. I wonder how many tokens that adds in though.
November 5, 2025 at 2:18 PM
TIL about datasette.io

Nice job, extremely useful
Datasette
Datasette is a tool for exploring and publishing data. It helps people take data of any shape, analyze and explore it, and publish it as an interactive website and accompanying API.
datasette.io
November 5, 2025 at 2:12 PM
Deep dark corporate secret: execs are often the target of phishing. A major HC concern was taken down for days, bare metal restore because of a chain of events started by a clueless exec clicking a link sent specifically to him.
November 4, 2025 at 12:01 PM
Suspect this too.
November 3, 2025 at 1:25 PM
Ive seen this too. 75% of the time that overdone response is kinda welcome, solves a problem you may have downstream, the other 25% of the time it just introduces problems you or Claude has to fix later. My next project will probably involve some front end instructions to keep things very simple
November 3, 2025 at 1:25 PM
That said, GitHub codespaces’ cloud env has been there for a few years and I think anthropic just borrowed that idea. It’s a necessary piece for this an now seems to be implemented very well by at least two of these players.
November 2, 2025 at 12:55 PM
Of course the very 1st thing I said when it attached to the repo : “read and understand this repo fully and summarize it”, then reviewed that to make sure it didn’t miss anything. It did not
November 2, 2025 at 12:46 PM
My main reason for doing this (starting GitHub codeshare and switch to cc) is to see how cc handled it and frankly it took one look at the repo and improved things immediately
November 2, 2025 at 12:39 PM
It has a planning, agentic actions, it fully reads and understands all readmes and markdowns in the repo before doing anything. What’s most useful for me is the personal Ubuntu cloud env it spins up to do all tests, etc in.
November 2, 2025 at 12:38 PM
You can choose the model just like the other Claude interfaces. You can even choose an external model that has an endpoint that supports cc, like GLM 4.6.

Honestly cc is really pretty slick.

bsky.app/profile/jooc...
Suggestion for anyone using claude code - consider changing the backend LLM to GLM 4.6 to save money. Close to as good as Sonnet 4.5, 1/3 the cost?

wsl_linux > alias | grep claude

alias cc='claude'
alias ccf='claude-glm-fast'
alias ccg='claude-glm'
alias ccg45='claude-glm-4.5'
November 2, 2025 at 12:35 PM
Pretty soon we’ll just all use “/usr/bin/aish” which will be “bash shell + universal terminal coding agent optional mode built in” and it will come as a standard shell like bash.
October 31, 2025 at 3:14 AM
It’s going to gel into a new better flow! maybe
October 31, 2025 at 2:31 AM