Brandon Rohrer

Robots and data and robots learning from data.
https://brandonrohrer.org 
#Robots
#Robotics
 #ReinforcementLearning 
#MachineLearning
#DataScience

Public Key

npub1xwxm0jud8zptjqv4zq82atmqjngewj2qghaa9g2m64d8mzf9vh0sjn4l9q

NIP-05 Address

brohrer@recsys-social.mostr.pub

Profile Code

nprofile1qqsr8rdhewxn3q4eqx23qr4w4asff5vhf9qyt77j59da2kna3yjkthcpz4mhxue69uhhyetvv9ujuerfw36x7tnsw43qz8thwden5te0dehhxarj94c82c3wwajkcmr0wfjx2u3wdejhgqgdwaehxw309aukzcn49ekk2tlzul4

Publishing to

relay.ditto.pub

nostr-pub.wellorder.net

yabu.me

Author Public Key

npub1xwxm0jud8zptjqv4zq82atmqjngewj2qghaa9g2m64d8mzf9vh0sjn4l9q

Show more details

Last Notes

2026-10-30 01:32:42 UTC

Listening to Patrick Stewart read sonnets tonight 😢

2026-10-26 21:43:10 UTC

Reuters posting a top-level headline about a 1% change in a poll with a 3% margin of error.
🧐
Trump approval rises from record low to 32% ahead of midterm election, Reuters Ipsos poll finds - https://www.reuters.com/world/us/trump-approval-rises-record-low-32-ahead-midterm-election-reuters-ipsos-poll-2026-10-26/

2026-10-13 15:31:19 UTC

Always surreal to visit the western US where strangers strike up conversations for no reason

2026-09-29 09:16:37 UTC

Optimizing models with lots of hyperparameters or functions with many options is a slog that usually requires a combination community wisdom, obscure blog posts, and tedious trial and error. Automated methods tend to either rely on strong assumptions of smoothness or brute force their way through every permutation. All of these options kind of suck.
I couldn't find a solution I liked to this problem so I built one: redsho, a robust evolutionary direction set hyperparameter optimizer.
https://cdn.masto.host/recsyssocial/media_attachments/files/117/353/549/299/773/090/original/a383a2b55fbbad83.mp4

2026-09-23 09:15:00 UTC

Impressive flex
#nevent1q…ec4g

2026-09-07 19:54:30 UTC

If you want something done right do it yourself
https://cdn.masto.host/recsyssocial/media_attachments/files/117/231/508/949/503/714/original/3a70e981c611c872.jpeg

2026-08-30 17:16:30 UTC

Million dollar idea:
Tater Toter
I’m not sure what it is. You’ll have to figure out design, manufacturing, marketing, supply chain, advertising, support, and compliance. So I figure we can split it 50-50.

2026-08-30 01:04:53 UTC

No marvel character is permanently dead until they get into a contract dispute with Disney

2026-08-25 00:37:48 UTC

Nothing like unwinding at the end of a long day with a steaming mug of mastodon. Bring out your cats!

2026-08-24 10:44:05 UTC

- reply

@nprofile…l2f6 I wish being right felt better than it does

2026-08-21 14:54:42 UTC

My team is looking for a Senior MLE, specializing in infrastructure.
We have a nicely balanced skillset, and perfectly sized *except* that we're missing this person:
https://jobs.ashbyhq.com/patreon/722d3541-d93a-4533-958b-74ce1060fb19
FWIW I'm having a great time and adore my team here. Come join us.

2026-08-20 01:03:42 UTC

If you’re into such things, I’m going to be chatting with Josh Starmer and Luis Serrano (two of the best ML educators I know) in a YouTube livestream, answering viewer questions, at 11:00am EDT Thursday (tomorrow/today).
8am PDT, 3pm UTC, 5pm CEST
https://m.youtube.com/live/Yx-ZkLftybY

2026-08-16 18:34:55 UTC

I hate that it’s necessary, but I celebrate the hell out of this project
https://www.climate.us/
https://cdn.masto.host/recsyssocial/media_attachments/files/117/106/623/693/650/106/original/24dbc389d6622e85.png

2026-08-13 13:26:49 UTC

Litter
#nevent1q…0z0h

2026-08-11 22:55:02 UTC

Elle Cordova is a wonder. If you'd like to tell her so, pop over to patreon and buy her a coffee.
https://www.patreon.com/cw/ellecordova
https://cdn.masto.host/recsyssocial/media_attachments/files/117/079/306/396/052/716/original/7812e5018c0c320c.mp4

2026-08-07 01:49:10 UTC

- reply

@nprofile…dyu0 don’t know if you’re looking for recs, but my wife tells anyone who will listen about The Menopause Manifesto. It is full of rage and bangers. https://www.goodreads.com/book/show/55272232-the-menopause-manifesto

2026-08-06 11:35:12 UTC

“Taste is built the slow, stupid, humiliating way: you make something bad, you are forced to live with it, it fails in front of you, and some part of you files the failure away. Then you do it again.”
https://notashelf.dev/posts/taste-is-all-thats-left

2026-08-05 11:45:27 UTC

- reply

@nprofile…vsjg possibly irrelevant thought: is the induced magnetic field through the center of the loop the manifestation of that radiation? Frequency of 0 (current is not oscillating) and 90 degrees out of phase (per Maxwell’s equations $?

2026-08-01 03:04:22 UTC

- reply

@nprofile…dyu0 according to my wife, this is very much the way

2026-07-14 00:08:06 UTC

If I do a really good life, then when I die there will be a dozen or so people that make a face and say “well fuck”

2026-07-11 08:43:51 UTC

Engineering is like sex.
If you optimize for speed you’re going to end up with unsatisfied customers.

2026-06-26 23:35:27 UTC

I shut down my online courses at End-to-End Machine Learning 18 months ago. Wrote a blog post and everything. But the platform I was using hasn't bothered to take them down. So they're just sitting there, all for free. Prego.
https://end-to-end-machine-learning.teachable.com
https://cdn.masto.host/recsyssocial/media_attachments/files/116/819/026/121/185/672/original/82026089457be110.png

2026-06-25 10:05:10 UTC

“The goal right now isn’t to optimize yourself into the shape the market demands. It’s to stay whole enough that when better conditions exist — and they will — you still have something left to bring.”
This piece may make you feel sad and angry and seen.
#nevent1q…mmzw

2026-06-25 00:36:32 UTC

My feed is 80% catsalad and I’m not mad about it

2026-05-12 11:37:10 UTC

I host my blog repo on Codeberg with mirrors on GitHub and GitLab. It started as a fun exercise in overkill, but I’m feeling better about it by the day. It’s three times the hassle. But maybe going multi-infra is worth the robustness benefit when trusted institutions start failing.

2026-05-11 11:39:10 UTC

- reply

@nprofile…yz28 kind of like how, once artificial diamond production got really good, the most valued diamonds are now the natural ones with small imperfections.

2026-04-28 11:14:44 UTC

Most of us have the “it can’t happen to me” blind spot. Human brains seem weirdly susceptible to delusion. And stubbornly reluctant to admit it.
#nevent1q…rn9h

2026-04-27 23:10:31 UTC

- reply

@nprofile…l2f6 awesome! I hope you rewarded yourself appropriately

2026-04-10 16:29:38 UTC

Hey #datascience people, new blog in which I aim to describe the complexities of being a Staff+ data science (and adjacent) roles. Let me know what I missed.
https://brandonrohrer.org/ds_roles
https://cdn.masto.host/recsyssocial/media_attachments/files/116/381/341/552/909/472/original/9eb7da1b5e3deb2c.png

2026-04-10 15:47:52 UTC

It's not just you--job search today really is a beast. As Melissa and Johnathan put it "the front door is fucked".
The secret? Focus on connecting with actual human people.
https://www.rawsignal.ca/newsletter-archive/500-applications-0-callbacks/

2026-04-07 09:52:25 UTC

This is what it sounds like when grown-ups talk about AI. Jonathan and Melissa deserve to be legends.
https://youtu.be/ncmuXQGGqBM?si=rH3zzbz5WwucA9Sf

2026-04-02 18:40:02 UTC

- reply

@nprofile…yz28 smells of word-by-word Google translate

2026-02-16 22:00:58 UTC

- reply

@nprofile…y4qv proof of human authorship

2026-02-03 23:29:22 UTC

As someone to whom "incurious" is one of the sickest burns imaginable, a tool that hands me code saying "it passes tests - don't worry about how it works" is about the least useful thing i can imagine.

2026-01-28 23:50:52 UTC

I wonder which is more damaging for you—everyone telling you you’re inadequate or everyone telling you you’re amazing

2026-01-21 11:33:36 UTC

https://cdn.masto.host/recsyssocial/media_attachments/files/115/932/870/203/753/483/original/8e42fa4e5cce2cdb.jpeg

2025-12-31 03:00:30 UTC

It is surprisingly hard to get computers to do addition with decimals the way a third-grader can.
adding floats gives things like
0.1 + 0.2 = 0.30000000000000004
https://graffitiwall.nexus/graffiti_adder.html
The mpmath library in Python is capable of arbitrary precision, but still gives things like
-1.2 + 1.21 = 0.009999999999...
https://mpmath.org
The answer I was looking for was Python's decimal library. It gives
0.1 + 0.2 = 0.3
and
-1.2 + 1.21 = 0.01
https://docs.python.org/3/library/decimal.html
https://cdn.masto.host/recsyssocial/media_attachments/files/115/811/941/266/725/311/original/e341c7ad07b83a88.png

2025-12-06 01:21:19 UTC

You’re not weak, you’re exhausted. Even the most badass dump truck stalls when it runs out of gas. Get you some gas.

2025-10-01 01:30:20 UTC

https://cdn.masto.host/recsyssocial/media_attachments/files/115/296/318/652/390/661/original/49045c6dbfb21402.png

2025-09-28 22:41:16 UTC

Caring for Your Webserver is out. It covers a grab bag of ways to get the most out of your self-hosted server.
- browsing the access logs
- catching missed pages
- automatically adding .html to requests, when needed
- redirecting URLs
- setting up log rotations
- finding a content provider for large files
Showing your webserver a little bit of love can speed up your website and cut down your bills.
https://brandonrohrer.com/hosting4.html
https://cdn.masto.host/recsyssocial/media_attachments/files/115/284/291/840/344/010/original/9658a0a43c8b9b10.png

2025-01-15 18:52:55 UTC

Either you die with your heroes or you live long enough to see them become the villains.

2025-01-02 12:17:01 UTC

I’ve always loved SQLite but now I love it more. Maintained by three people. 600 lines of test code for every line of code. Design decisions favor robustness over feature richness at every return. However widely you think it’s used, your guess is way low.
A striking example of the weird and wonderful things that can be built when shareholder value is not the primary goal.
https://fosstodon.org/@amoroso/113758608409377815

2024-12-26 13:45:50 UTC

- reply

@nprofile…8f02 This is really good. Thanks for this.

2024-12-24 21:33:27 UTC

I've been trying to learn about HTTP clients in Python, so I wrote about HTTP clients in Python.
https://www.brandonrohrer.com/http_client
If you notice any mistakes please point them out! I hid three mistakes in there on purpose to make it interesting, but if you find the fourth I will pay you five Internet Points.

2024-12-23 22:36:23 UTC

Them: What side projects are you working on over the break?
Me: ummm...a deep dive on HTTP status codes?
https://cdn.masto.host/recsyssocial/media_attachments/files/113/704/524/682/624/538/original/1834a251499ee898.png

2024-12-23 19:54:13 UTC

The siesta is is the crowning achievement of human civilisation.

2024-12-22 18:40:56 UTC

I will pay you ten thousand dollars for a fifteen minute nap

2024-12-02 02:39:00 UTC

I'd like to introduce dsmq, the Dead Simple Message Queue, to the world.
Part mail room, part bulletin board, dsmq is a central location for sharing messages between processes,.
Its defining characteristic is bare-bones simplicity.
www.brandonrohrer.com/dsmq
🧵

2024-11-24 01:02:06 UTC

Adding my love letter to
arxiv.org/pdf/2304.01315
Empirical Design in Reinforcement Learning
by
Andrew Patterson, Samuel Neumann, Martha White, Adam White
JMLR 25 (2024) 1-63
#ReinforcementLearning
These aren’t the heroes we deserve, but they are the heroes we need.

2024-11-19 12:04:33 UTC

in case you need to hear this
again
it’s ok to leave people
even your blood
it’s ok to find new people
it’s ok to have no people
there’s no wrong way to go through life
only ways that hurt you
and ways that heal

2024-11-18 11:59:16 UTC

I’ve been loving Bluesky these last couple weeks because I missed my old Twitter friends so much. I wondered whether it would replace my Mastodon experience but now I can confidently say that it won’t.

2024-11-11 11:47:18 UTC

“Good Judgment depends mostly on experience and experience usually comes from poor judgment.”
—This was funnier when it wasn’t happening on a global scale

2024-11-07 14:34:59 UTC

In case you’re wondering who wants to hear about your dorky side project, it’s me. I do.

2024-08-17 11:31:42 UTC

- reply

@npub1sys…ckr3 I go this route by default too. I mean, if you want a strongly typed language there are plenty to choose from right?

2024-06-12 12:24:46 UTC

Going to popularize working end to end projects with a small collection of lightweight tools as being a “short stack engineer”

2024-06-07 13:19:40 UTC

- reply

@npub1xr8…3q0g nothing a little session at the climbing gym won’t heal!

2024-06-02 11:45:23 UTC

One danger of letting employees take off several weeks at a time is that they remember what it feels like to be happy.

2024-05-27 13:25:14 UTC

The worst thing about writing your own tools is sending frustrated bug reports to your own inbox.

2024-05-10 23:49:41 UTC

Hofstadter's Law but for p-values:
You probably don’t understand p-values as well as you think you do, even after taking this law into account and studying up on p-values.
Inspired by @npub1wky…6c6c ‘s banger
https://mastodon.social/@kjhealy/112418946802530131

2024-05-08 11:52:46 UTC

- reply

3. Without carefully boxing in how an AI system is deployed there is always a long tail of special cases that take decades to discover and fix. Paradoxically all those fixes are AI-complete themselves.
3/3

2024-05-08 11:52:29 UTC

- reply

2. Most successful AI deployments have a human somewhere in the loop (perhaps the person they are helping) and their intelligence smooths the edges.
2/3

2024-05-08 11:51:54 UTC

Reposted from @npub1cep…af86 on LinkedIn:
About a month ago I posted my three laws of robotics. Here are my three laws of Artificial Intelligence.
1. When an AI system performs a task, human observers immediately estimate its general competence in areas that seem related. Usually that estimate is wildly overinflated.
1/3

2024-05-02 13:27:18 UTC

- reply

@npub1sys…ckr3 please tell me you felt the uncontrollable urge to shout “get off my lawn!”

2024-04-22 16:17:45 UTC

Seven stages in the professional development of an engineer:
1. I’m Kirk!
2. I’m Spock.
3. I’m Scotty.
4. I’m Bones, dammit.
5. I’m Troi?
6. I’m Picard.
7. Why am I wearing a red shirt?

2024-04-21 15:27:59 UTC

Me, a data scientist brought in ten years ago to automate expensive matter experts out of a job: look man, it’s all in the name of efficiency
Me, being replaced by a bot: not like that

2024-04-20 11:21:48 UTC

This is a great trick for me, someone who Points Out Problems as a Service
https://www.simplermachines.com/why-you-need-a-wtf-notebook/

2023-10-19 21:52:01 UTC

Do you like stories? I have a story.
1/
As the oldest grad student in my lab along time ago, I was also the default system administrator. None of us knew anything about system administration, but I was the lucky one tasked with sorting it out.
Our university administrators, in their wisdom, decided to levy a tax on research labs to support the university's recently installed ethernet cable network. They settled on $25 per month per IP address.

2023-09-01 17:37:32 UTC

- reply

@npub1960…w6j3 maybe so, but I find photos of Buttercream quite therapeutic

2023-08-18 00:11:24 UTC

- reply

@npub1efv…4jx2 can one die from an overdose of irony? Asking for a friend