Sign in

yakhu

@gremble.xyz
20 followers 29 following 240 posts

Mathematician | Computational Scientist | Uncompetitive competitive weightlifter. Into synths, programming and whatever else tickles my fancy. www.gremble.xyz he/him/his

PostsRepliesMedia
yakhu @gremble.xyz · 05/10/2026
We are hiring and it is wild how many people with accounting degrees have switched to software development here.
020
Reposted by yakhu
Angela Collier @acollierastro.bsky.social · 01/10/2026
I just think women should be able to get educations without a constant threat of sexual violence.
91382206
yakhu @gremble.xyz · 01/10/2026
Fokken hond. All relationships are long distance relationships now.
031
yakhu @gremble.xyz · 30/09/2026
SQLGlot 10/10 library.
001
yakhu @gremble.xyz · 26/09/2026
When one of your citations is so fresh off the presses, the paper is still warm: ieeexplore.ieee.org/abstract/doc...
ieeexplore.ieee.org
Hierarchical Learning-Based Control for Multiagent Shepherding of Stochastic Autonomous Agents
Multiagent shepherding represents a challenging distributed control problem where herder agents must coordinate to guide autonomously moving targets to desired spatial configurations. Most existing co...
000
yakhu @gremble.xyz · 26/09/2026
It should be obvious that I am working through my procastrination. I've made a diagram. The keys changed because the staggering was degrading performance more than I wanted.
000
yakhu @gremble.xyz · 26/09/2026
In the spirit of Le Cunn's talk, I am adding MPC into my ablation study. "What if we entirely ablate the RL step and use a 2-step look-ahead MPC controller?" The neat part of that is that I get to do some more math.
000
yakhu @gremble.xyz · 26/09/2026
Nothing works, but at least I have pretty plots.
110
yakhu @gremble.xyz · 25/09/2026
I've built my little guy to touch the keys. It can tell time to do a beat. I implemented randomised euclidean rhythms and random notes in a fixed key. Now the player won't adapt to different conductors. Yay for getting to report a negative result :|
000
yakhu @gremble.xyz · 21/09/2026
It is always fun to do computational experiments where the reward function suddenly collapses after you made a change you thought was good :D
100
yakhu @gremble.xyz · 20/09/2026
Implementing numerical algorithms from a paper is quite difficult. I was pleased to get this working last night. Now I just need to figure out how to add another coupling equation to model the diffusion of the chemo-attractant over the graph.
100
yakhu @gremble.xyz · 17/09/2026
At least there is a highveld thunderstorm cracking overhead
000
yakhu @gremble.xyz · 16/09/2026
I just want to know what is the minimum amount of information you need to share for two little guys to be able to work together. Now I need to do Lebesgue integration :(
010
Reposted by yakhu
Timnit Gebru @timnitgebru.blacksky.app · 14/09/2026
The only person in public service who has consistently made 100% sense on issues of "AI," which is why they hate her so much.
Law enforcers already have authority to charge companies and their CEOs for creating and releasing dangerous, unvetted, or defective products. We shouldn’t let discussions about new legal regimes distract from the fact that there’s no AI exemption from laws already on the books — a point 
@FTC
 emphasized repeatedly during my tenure.

1. There is an extensive set of laws that govern dangerous and defective products. For example, releasing unvetted AI models or agents can violate consumer protection laws. Shipping flawed AI tools without implementing adequate measures to detect and stop rogue or defective AI agents can be an “unfair or deceptive” act or practice under the FTC Act (and analogous state laws). And some state AGs are already exploring holding AI firms and their CEOs criminally liable when their models participate in criminal activity.

2. Existing laws also prohibit “unfair methods of competition.” This covers instances where AI firms appropriate the competitively sensitive information of their customers, including through tracking their use of various tools. It can also cover instances where firms pursue dangerous behavior, aware that doing so may compel rivals to do the same.As the Supreme Court has noted: “A method of competition which casts upon one's competitors the burden of the loss of business unless they will descend to a practice which they are under a powerful moral compulsion not to adopt, even though it is not criminal, was thought to involve the kind of unfairness at which the [unfair methods of competition] statute was aimed."

3. The highly concentrated and interconnected structure of these markets could be creating major risks and conflicts of interest. We had started investigating these partnerships and cross-investments across the stack (and released a preliminarily overview of some findings: https://ftc.gov/news-events/news/press-releases/2025/01/ftc-issues-staff-report-ai-partnerships-investments-study). 

Both federal and state enforcers should be scrutinizing these opaque relationships and inter-dependencies. We are already seeing how these relationships could undermine accountability. For example, OpenAI could face liability given the Hugging Face incident, but Hugging Face being bought up by Nvidia means that we’re unlikely to see it file a lawsuit over this — given Nvidia’s strong incentive to see OpenAI continue full speed ahead.4. As AI tools dramatically change the landscape of cybersecurity risks and hacks, all businesses should be doubling down on having core security protections in place. Firms that fail to invest in adequate data security measures or fix known vulnerabilities can also be breaking the law. A recent analysis showed that around 1/3 of Fortune 100 companies do not even have a way to notify them about security issues. During my 
@FTC
 tenure, we sued firms for poor data security practices and held CEOs liable when they were personally responsible.

https://this.weekinsecurity.com/dozens-of-americas-largest-companies-have-no-simple-way-to-report-security-flaws/

https://ftc.gov/news-events/news/press-releases/2022/10/ftc-takes-action-against-drizly-its-ceo-james-cory-rellas-security-failures-exposed-data-25-million

5. As policymakers consider new legal regimes, we should be looking to lessons from prior efforts to govern major sectors, such as banking and other networks, platforms, and utilities. Tools like structural separations, nondiscrimination, and supervision could be key, and there’s a rich history of what works and what doesn’t. But we can and must pursue any new efforts alongside enforcing existing laws.
192026658
yakhu @gremble.xyz · 12/09/2026
It is so demeaning to have to admit that my research is in AI. I am doing marl research just because I want to figure out how to get the little guys to play games effectively. I just want to know how to make decisions when working together haha
000
yakhu @gremble.xyz · 10/09/2026
I should not be allowed to make decisions. I am working on a project computation differential equations and I realised that I am making my group members do an entire additional honours project for this subject. I am teaching them methods to solve DEs on graphs.
000
yakhu @gremble.xyz · 09/09/2026
If we put the amount of compute going into the pattern recognition machine through almost any other system, I am fairly sure that we would be able to develop spectacular results from that as well.
011
yakhu @gremble.xyz · 05/09/2026
If you watch one video today, it should be this www.youtube.com/watch?v=d4Yd...
youtube.com
Maybe there is a Category of Categories
YouTube video by Sheafification of G
000
yakhu @gremble.xyz · 02/09/2026
I love living in a place where I can walk to a restaurant. I really wish that for everyone. Mixed use neighbourhoods need to be more common
000
yakhu @gremble.xyz · 02/09/2026
I am so tired of reading AI shit.
010
yakhu @gremble.xyz · 28/08/2026
I really need to be preparing for a fluid dynamics test, but I found [1]. I really cannot be trusted. [1]: math.stanford.edu/~vakil/216bl...
math.stanford.edu
010
yakhu @gremble.xyz · 28/08/2026
Just so we are all aware: I am getting paid to build bayesian belief models on fiber structures. It is exciting, but I am not telling them I am doing that. They think I am doing data engineering :)
010
yakhu @gremble.xyz · 27/08/2026
Working for a startup is really stupid shit.
000
yakhu @gremble.xyz · 27/08/2026
Good. The validation that I needed today.
000
yakhu @gremble.xyz · 27/08/2026
It is really awkward when you built a really nice and neat collocation method and then you forget to add a term to cleanly handle a non-homogeneous equation. It only clicked how to account for it on the walk to my car, so now I just won't get those marks but at least I know how to solve the problem
000
yakhu @gremble.xyz · 24/08/2026
"Libertarian societies naturally emerge on a fluid" is a concept in fluid mechanics that we haven't gotten to yet. It is probably after we deal with turbulence. youtu.be/C5ix9rcLpPM?...
youtu.be
no one wants to live here
YouTube video by 2 Danny 2 Furious
000
Reposted by yakhu
Dr. Fim @fim.bsky.social · 22/08/2026
Ansel Adams on assignment for IBM - one of my favourite shots. Not the usual landscape but the skill of a woman weaving magnetic core memory. The beat up nails captures total “woman in STEM” vibes.
A woman with extremely short painted nails weaving magnetic core memory. A series of rows and columns of wires in a grid.
252936873
yakhu @gremble.xyz · 21/08/2026
Spent my evening learning about pseudo-spectral methods for solving ODEs and PDEs. I figured out vandermonde matrices, now I just need to figure out how to get all of this to work for abitrary order DEs before Wednesday :D
000
yakhu @gremble.xyz · 14/08/2026
"Very smart people have already defended humanities scholarship for its difficulty as well as its sometimes intimidating specialization." This tickles me so much because mathematics doesn't seem to need this defense. defector.com/humanists-ha...
defector.com
Humanists Have Done Nothing Wrong | Defector
For many years, I’ve held a pet theory that cultural consumers can be grouped into two broad categories: Primary Source People and Secondary Source People. The former are the book lovers, cinephiles, ...
000
yakhu @gremble.xyz · 12/08/2026
Today was a good research day. I discovered that the quantity I am looking for has a name: Exformation. I also learned about Blackwell's Informativeness Theorem and the formulations surrounding that. So now I think I have a better language to describe my problem. Neat :)
000
yakhu @gremble.xyz · 12/08/2026
I fucking love a dried fruit cylinder.
A bowl of dried fruit cylinders
000
yakhu @gremble.xyz · 10/08/2026
Imagine taking commentary from the Automobile Association of South Africa on planned work on a rail network. Obviously they will give reasoned and unbiased commentary...
000
yakhu @gremble.xyz · 06/08/2026
What do you mean I need to figure something out and not just know it intrinsically? Also, Information Theory is teaching my that my intuition is worth shit.
010
yakhu @gremble.xyz · 06/08/2026
It is really fun that "crimes as marketing" is now the trend for AI companies.
000
yakhu @gremble.xyz · 27/07/2026
Being a data engineering expert and billing time as a data engineering expert really sucks when you then have to apply your data engineering expertise to engineering data.
100
yakhu @gremble.xyz · 27/07/2026
ODEs keep humbling me. The fact that I cannot solve an ODE at the drop of a hat is something that I should work on. At least the easy ones - I am willing to give myself a pass on the beefy non-linear ones.
010
yakhu @gremble.xyz · 25/07/2026
First time snatching in months after hurting my wrist (ganglion cyst). I worked up to 50kg and did three sets. I feel crusty and out of practice, but my coach think they look nice so I will take it.
020
Reposted by yakhu
Dr Heidy Khlaaf (هايدي خلاف) @heidykhlaaf.bsky.social · 22/07/2026
The coverage on this OpenAI incident is abysmal. Use of the terms "rogue" and "loss of human control" lead to groupthink as people lack the critical skills to understand the difference between "autonomy" and faulty reward functions in AI on a task it was directed and given access to do.
217870
yakhu @gremble.xyz · 21/07/2026
Today was simplifying approximations of the navier-stokes equation in curvilinear coordinate systems. Actually just boundary layer things using Prandtl's approximations, but wow - I am definitely still a bit rusty.
000
yakhu @gremble.xyz · 20/07/2026
They tricked me. Mathematical modelling is just deriving and solving PDEs. I've been hoodwinked.
000
yakhu @gremble.xyz · 15/07/2026
Studying full-time after almost 10 years has gone more successfully than I had hoped - with geometry of PDEs being a notable exception 😅. However, I got two distinctions (research and symmetry methods in differential equations) and two Bs for robotics and global optimisation. I can **still** do math
110
yakhu @gremble.xyz · 10/07/2026
I think I just built a multi-tenant datalake. It uses ducklake and postgres to run the catalog. My query execution happens using isolated processes running ducklake in-memory databases. And I have full tenant data isolation. I expose a postgres-like dsn so psql can just connect to it.
120
yakhu @gremble.xyz · 04/06/2026
If I didn't have so many dependents, I would probably take a pay cut to get to make things with my hands. I am tired of fighting with software.
000
yakhu @gremble.xyz · 30/05/2026
An absolutely pointless bit of wisdom: An engineering course is much easier if you already have a degree in mathematics.
000
yakhu @gremble.xyz · 20/05/2026
I am once again faced with my old nemesis: writing down what I know.
000
yakhu @gremble.xyz · 19/05/2026
When you run out of symbols so you just use SQUARE
000
yakhu @gremble.xyz · 19/05/2026
I am really grateful that for a transversally foliated neighbourhood of a manifold with dimension m, I can identify the projection along the fibres \pi with the normal bundle.
000
yakhu @gremble.xyz · 17/05/2026
I am not having fun with ROS and SLAM :)
000
yakhu @gremble.xyz · 17/05/2026
I am busy doing differential topology, specifically the symmetries and conservation laws of partial differential equations. It feels like I am holding an increasingly larger set of balloons that I am carefully using to do things but if I let one go, all of them will go.
100