Good writeup about SI authoring experience and how it shortens feedback loop in infra development

https://matthewsanabria.dev/posts/take-the-system-initiative/

Founders mode
Paul Graham wrote handwavy and vague article https://paulgraham.com/foundermode.html and introduced dichotomy of managerial style vs founders style - without defining it. It lead to discussions in industry, here is few good writeups https://skamille.medium.com/founders-create-managers-aba3c88981ba https://oxide.computer/blog/reflections-on-founder-mode

And a few links mentioned in them https://www.oreilly.com/radar/how-i-failed/
https://threadreaderapp.com/thread/1216491216356823040.html

And that reminded me of Wardley Maps Doctrine that defines what is needs to be done to set company wide culture and set universally applicable principles.
https://www.wardleymaps.com/doctrine

Panel discussion about modern IaC https://fixupx.com/PulumiCorp/status/1836433750428647466?t=LcJB6KGvwnudSv26hXu_qQ&s=19

Another essay by Mark Burges
"I once called microservices “a temporary aberration in the history of computing” because they are really a human workaround for a problem that has typically been solved using compilers in other cases, e.g. Remote Procedure Calls (RPC)."

It's a special pleasure (as a physicyst myself) reading text on intersection of physics and configuration/data management and cloud

https://mark-burgess-oslo-mb.medium.com/the-sorrow-of-the-cloud-messengers-fb0c43de27f3

What happens when you focus on user needs
https://blog.gitbutler.com/why-github-actually-won/

Finally password rules from NIST that - does not require periodic changes
https://fixupx.com/blackroomsec/status/1838371397590286677

System inititiative launched general availability today - so you can try it for yourself
https://www.systeminit.com/blog

https://youtu.be/20vIl7I0R4I

Another bookmark
Article about writing code for Humans

https://erikbern.com/2024/09/27/its-hard-to-write-code-for-humans

Great deep dive into H100 pricing

In summary, for an on-demand workload per h
>$2.85 : Beat stock market IRR
<$2.85 : Loses to stock market IRR
<$1.65 : Expect loss in investment

https://www.latent.space/p/gpu-bubble

Metas vision:
rack-scale design
up-to 140kW per rack
400 VDC unit
24k GPU cluster

https://engineering.fb.com/2024/10/15/data-infrastructure/metas-open-ai-hardware-vision/

and very deep dive into datacenter electrical anatomy

https://www.semianalysis.com/p/datacenter-anatomy-part-1-electrical

Good interview with Gwern, author of gwern.net
https://www.dwarkeshpatel.com/p/gwern-branwen#§transcript

A brief story why Pat Gelsinger was wrong for Intel and how that affected hardware decisions at oxide

https://bcantrill.dtrace.org/2024/12/08/why-gelsinger-was-wrong-for-intel/

View on AI from power demand point of view, predictions and estimations of future needs.
All major players are buing nuclear, and had to invest in green
Which wins: Koomey’s Law or the Jevons effect?
https://about.bnef.com/blog/liebreich-generative-ai-the-power-and-the-glory/

Nice google Zurich tradition of hacking password of leaving collegue

https://bughunters.google.com/blog/6355265783201792/the-great-google-password-heist-15-years-of-hacking-passwords-to-test-our-security-and-build-team-culture

A bit of history of Wardley mapping and ubuntu:

"Well, the process roughly went ...

1) A friend of mine suggested to Mark that he hire me as their new head of design.
2) I turn up to a meeting on Friday, Mark goes "Apparently you're a great designer"
3) I respond ... "I know nothing about design, I'm a lousy designer"
4) ... silence
5) ... more silence
6) Mark says "Well, I've got 30 minutes, anything you want to talk about"
7) I talk mapping and cloud
8) 3.5 hours later Mark hires me to run strategy / cloud for Canonical. Starting Monday. This was 2008.
9) First five meetings at Canonical followed the same format ...
X: "What are you here for?"
Me: "To help Canonical and Ubuntu get into the cloud"
X: "Cloud is a fad, we should focus on the enterprise"
10) Each of the meetings I used maps to pick up a few allies. Soon enough there was a core group including Rick Clark and Soren.
11) We used the maps to identify where to focus - we knew there would be new practices (not what they would be called), new things built on cloud (not what they would be). We went out of our ways to help others build on Ubuntu. We also stopped groups in Canonical hindering the community development / use of Ubuntu.
12) 18 months later, for the cost of $500k ... we had gone from 3% of the operating system market to 70% of all cloud."

A very long read on why nvidia shares were traiding low in past 3 days
-deepseek
-competitors creating their chips
-tinygrad
https://youtubetranscriptoptimizer.com/blog/05_the_short_case_for_nvda

https://www.calcalistech.com/ctechnews/article/s1tra0sfye

How intel failed to utilize its AI investments:
- didnt risk buying mellanox
- bought Habana labs and put it under bureucracy burden and mismanagement
- didnt commit to a single path and spread efforts in multiple AI projects

Wallbleed: A Memory Disclosure Vulnerability in the Great Firewall of China

https://gfw.report/publications/ndss25/en/

Great article on how-to use llm for code,
you need to switch between vibe coding and dictator in order to get results, and context is a king
https://simonwillison.net/2025/Mar/11/using-llms-for-code/

Deep dive into cpu memory latency

https://chipsandcheese.com/p/pushing-amds-infinity-fabric-to-its