RE: https://chaos.social/@delta/115842446685941927

Encrypted content / headers / transport / etc. are great harm reduction, but by no means is it metadata avoidance (in the strict academic sense of what those words mean when applied to communications systems).

Third party actors (e.g. relay servers / network actors) can build correlative models using the *metadata* inherent in the communication protocol (source, destination, timing, frequency).

We have decades of research demonstrating the power of such models - let's not minimize them.

This is an excellent introduction to a modern Machine Learning stack, discussing how PyTorch, vLLM, Kubernetes, and Anyscale Ray fit together. I thought the talk did an excellent job motivating the need for a distributed computing layer (Ray) optimized for complex AI applications, including both training and inference.

I think it is inspiring that Robert and Ion Stoica started building Ray to scratch their own itch.

https://youtu.be/JEM-tA3XDjc by Robert Nishihara via PyTorch Conference 2025

@jeffjarvis Looking forward to watching this!

ps. Something funny seems to be going on with this channel, the last few episodes seem to have duplicate uploads

@drmorr "Be your own boss," they said. "It'll be great," they said.

Happy new year, Dr Morrison. I'm looking forward to more nerdy antics from you this year. 💖

@irene @dev @palvaro @dan Dan is getting the real gift from the looks of this soup 😍

@drmorr I bet it's because your boss really cracks the whip, huh?

"We measure cybersecurity exposures by simulating port scan
attacks and counting the number of publicly accessible vulnerable ports for Fortune 500 firms. We show that high exposure firms exhibit negative excess returns relative to low exposure firms. The estimated underperformance amounts to 0.33% per month, or approximately $75mm in value
lost for the median size firm in the Fortune 500."

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3746589 by Christos Makridis and Tim Liu via SSRN

"Harder-to-automate skills older workers picked up during their careers might be insulating them from the same kind of AI hit [that early-career people are facing]. A senior software developer, for example, might have learned how to work collaboratively with noncoders and deliver the product the company needs. Such skills remain highly valued to employers, and may never be automated away."

https://www.wsj.com/economy/jobs/ai-entry-level-job-impact-5c687c84 by Justin Lahart via The Wall Street Journal

@saraislet Hm, can you elaborate?

I'm not sure where the fraud is here? Or what you mean about the finance industry.

I'm genuinely curious, not being confrontational here. I think there are many things involved that people consider varying degrees of unethical :)

I will confess that I am envious of the self-awareness that Isabel has at her age. And I will admit that, even with the wisdom that my age has presumably granted me by now, I am grateful for this reminder.

"So much of life is falling off the path and then starting again.

And I think that's okay. And it's part of the journey of life. Choosing over and over, what we give our energy to, and letting go of the things that no longer serve the person we're becoming."

https://youtu.be/bhWi9mELOgU

"Several decades ago, the sociologist Erving Goffman identified unemployment as a 'spoiled identity.' What he meant was that the unemployed are denied full participation in social life because others view them with suspicion."

https://hbr.org/2020/04/when-losing-your-job-feels-like-losing-your-self by Aliya Hamid Rao via Harvard Business Review

"So, controlling for occupation, I found that the returns to double-majoring in liberal arts and STEM were 5.2 percent, and 3.4 percent with a business degree. In other words, even when we look within narrow occupational categories, those who double-majored across fields tended to earn more than those with a single degree."

https://theconversation.com/does-it-pay-to-get-a-double-major-in-college-74420 by Christos Makridis via The Conversation U.S.

@saraislet What do you mean? Are you saying that churning (routinely signing up for new credit cards for the signup bonus) or optimizing credit card points (by using different cards for different categories of purchases) is unethical?

A fantastic survey of open source model training/inference architectures. I liked that the vignettes explain how vLLM, Anyscale Ray, PyTorch, and Kubernetes work together. The diagrams and tables are superb -- I thought it was really great to see how world-leading organizations like Pinterest, Uber, and Roblox adapted their machine learning stacks over time.

https://www.anyscale.com/blog/ai-compute-open-source-stack-kubernetes-ray-pytorch-vllm by Robert Nishihara via Anyscale

This is a fantastic short read about finding security vulnerabilities in open source systems using static analysis tooling. I liked the approach of using scripts/automation to systematically explore different payload configurations.

https://blog.trailofbits.com/2025/08/04/uncovering-memory-corruption-in-nvidia-triton-as-a-new-hire/ by Willis Vandevanter via @trailofbits

Big updates in Gradle 9, including:

1. Dependency verification: Checksum and signature checking for dependencies.

2. Performance improvements for IDE integration: Tooling API has up to a 12% speedup for large builds with many up-to-date tasks.

3. Reproducible builds by default: Byte-for-byte identical content for JARs, ZIPs, and other Java packages, along with consistent timestamps, permissions, and file ordering.

https://gradle.org/whats-new/gradle-9/ via Gradle

"Redditors are frequently sophisticated with their spreadsheets; many of them could clearly earn three orders of magnitude more from the financial industry if they stopped thinking that the right way to monetize spreadsheet skill was in gaming credit card signup bonuses.

The biggest difference is that you’re optimizing over portfolios, over time and Redditors are largely playing in single player mode (and frequently over short horizons)."

https://www.bitsaboutmoney.com/archive/anatomy-of-credit-card-rewards-programs/ by Patrick McKenzie

"People we invited to embrace discomfort were more motivated: They persisted longer in improvisation exercises, engaged more in an expressive writing exercise, and opened themselves up to challenging but important information. Personal growth is sometimes uncomfortable; we found that embracing discomfort can be motivating."

https://journals.sagepub.com/doi/10.1177/09567976211044685 by Kaitlin Woolley and Ayelet Fishbach via Association for Psychological Science

"New cyber insurance claims data from Resilience’s portfolio illustrates the financial fallout of this domino effect, finding that third-party risk, including ransomware and outages affecting vendors, accounted for 31% of all claims in 2024. Even more startling, third-party risk led to claims with incurred losses for the first time ever, making up nearly a quarter (23%) of incurred claims in 2024 (compared to 0% in 2023)."

https://cyberresilience.com/cybersecuritys-biggest-blind-spot-third-party-risk-new-resilience-analysis-finds/ by Whitney Glockner Black via Resilience

A thoughtful look at money and our values, as told through three vignettes of lottery winners.

How are we spending our time and money budgets to acquire life's real luxuries: quality time with others, health, sleep, and guilt-free relaxation? What matters to you?

https://youtu.be/L2OVDJ6cDJc by Matt Pitcher via TED Conferences