Glad to see Meta is finally not censoring conservitive content. /s

Vim: set shiftwidth=4 smarttab

We can afford to just use spaces now

There are benifits to working on the same campus. Though the problem with trying to set up office overseas is now you have to flex to their cultural expectations have work with their legal frameworks. You may have to navigate their bribery/lobbying system as well, which might require even more overhead. All of which duplicative if they have to have that somewhere else in order to leverage the much more lucrative investment markets.

These are for profit companies the general motivation for them to do anything is to maximize the extraction of value. You’re right, they aren’t anti American nor pro immigrant. In a different geopolitical and economic market they would gladly pay warlords to maintain an active slave population as long as doesnt personally effect their social lives and the profit margins were wider then the h1b1 program.

Tab gang

   Rise
     Up

Huh, a kind of slopbot glory hole system

Personally I like the exception message and debug message and log outputs approach.

Comment just for what you said, a kind of historical graffiti.

Exceptions should catch why something didn’t work and where. Debug should be telling where in the process something is happening. If it’s descriptive enough without code in front of me it extra useful with code in front of me.

ADRs, and tests should cover the whys and expectations for a code base.

I do wish I had a better structured framework for debug statements that could be tested/verified easier (do you all right tests for debug outputs? That sounds like hell tbh).

Is there a way to read the article that doesn’t download a file?

Everyone has said Openscad because it’s cool but Blender also has a python API fyi!

“AsCode” is so freaking neat to me. Mermaid.js make diagrams as code. Openscad so you can CAD as code. LaTex for making writing mathmatical formulas (and more it’s a very rich mark up language). Reveal.js to make a slideshow as code. Markdown can make some really nice documentation. With tools like Markdeck so you turn that into a slide deck.

The infra as code and config as code (k8s yaml, helm, open tofu, ansible, Guix, nix, etc) are also super powerful but are kind of just tools for computer people. Though some of policyAsCode stuff seems like it can start to blur some interesting lines too, either by putting weird things k8s, or in your ci/cd workflow or by OPA/cedar integrations.

There is also digital twins! Eclipse BaSyx is apparently a growing standard for but that is a neat thing creating really world system models that you can run simulations on but also run real world equipment off of to.

Also apparently (typing this got me interested to see what was new out there!) there is OpenFisca for modeling tax policy as code that can be used to simulate how tax code would effect households! Catala is a “laws as code” DSL for turning legal texts into executable programs. Also smart contracts, and Decentralized Autonomous Orginizations (DAOs) in the crypto space are neat as a tech (ignoring the grifting in the space since they got linked to US investment market).

Is there really a market for mass production of low quality software?

Maybe it’s just never been needed in any of the market segments I work, but it always seems like it’s better to spend 40 hours today to refactor or clean up code then to spend 10000 hours bug hunting spaghetti code. Like sure your perl script worked but for the love of all let’s actually make a project with a maintainer and a code base you can reasonably attach an SLA to plz

Home and family: Has working nodes

Hell yeah, I’ll take that assumption

Not that I see either looking at opensource inference engines. vLLM supports setting the seed so if you have that you do, and maybe it’s just as simple as harness setting and recording that. At least for replay this request type of operation but that does not show where in model each random numbers would have been applied.

Me too

We are saying the samethings but you are adding no to it.

Right, during inference random numbers are generated, that plus the numbers from input are added to the weight values and the matrix multiplication happens. If you used the same random numbers and inputs it is deterministic. For regular use you don’t do that because you want a stochastic output, if you wanting to do forensics and trace what led to an output you would benifit from that determinism.

This is one of my critics of wilderness areas (I love them otherwise). Without having a place for people to be apart of and with the rest of the natural ecology of an area it’s too easy to see avoiding active harm as an optional charity instead of a needed part of maintaining a living biosphere

Tbf this walking just with robot leg support

Well a Koch brother is. The other one is trying to privatize firey pits if the lore is to be believed.

Honestly for physical therapy if these are built right a huge amount of people could benifit from this. For people with no treatment options too that just need assistive tech.

I mean it is a semantic collection of weights tied to tokens (fractions of written language). There is also tool calls which could be analysied for additional input analys.

No that should be discoverable with the models weights, input and random numbers added to the weight at the time.

Say for example you find that a collection of outputs behave oddly or in an undesired way, you could use this to find what simularties they share with each other but delta with other and naively prune the nodes or simply decrease their weights. Those you could also try to corralate that to certain input tokens to engineer better prompts or try to trace it back to initial training data.

It could also be a failed tool call adding garbage data in, or malicious. A trace could catch that as well.