frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

fp.

The original vi is a product of its time (and its time has passed)

https://utcc.utoronto.ca/~cks/space/blog/unix/ViIsAProductOfItsTime
1•ingve•6m ago•0 comments

Circumstantial Complexity, LLMs and Large Scale Architecture

https://www.datagubbe.se/aiarch/
1•ingve•13m ago•0 comments

Tech Bro Saga: big tech critique essay series

1•dikobraz•16m ago•0 comments

Show HN: A calculus course with an AI tutor watching the lectures with you

https://calculus.academa.ai/
1•apoogdk•20m ago•0 comments

Show HN: 83K lines of C++ – cryptocurrency written from scratch, not a fork

https://github.com/Kristian5013/flow-protocol
1•kristianXXI•25m ago•0 comments

Show HN: SAA – A minimal shell-as-chat agent using only Bash

https://github.com/moravy-mochi/saa
1•mrvmochi•25m ago•0 comments

Mario Tchou

https://en.wikipedia.org/wiki/Mario_Tchou
1•simonebrunozzi•26m ago•0 comments

Does Anyone Even Know What's Happening in Zim?

https://mayberay.bearblog.dev/does-anyone-even-know-whats-happening-in-zim-right-now/
1•mugamuga•27m ago•0 comments

The last Morse code maritime radio station in North America [video]

https://www.youtube.com/watch?v=GzN-D0yIkGQ
1•austinallegro•29m ago•0 comments

Show HN: Hacker Newspaper – Yet another HN front end optimized for mobile

https://hackernews.paperd.ink/
1•robertlangdon•30m ago•0 comments

OpenClaw Is Changing My Life

https://reorx.com/blog/openclaw-is-changing-my-life/
2•novoreorx•38m ago•0 comments

Everything you need to know about lasers in one photo

https://commons.wikimedia.org/wiki/File:Commercial_laser_lines.svg
2•mahirsaid•40m ago•0 comments

SCOTUS to decide if 1988 video tape privacy law applies to internet uses

https://www.jurist.org/news/2026/01/us-supreme-court-to-decide-if-1988-video-tape-privacy-law-app...
1•voxadam•41m ago•0 comments

Epstein files reveal deeper ties to scientists than previously known

https://www.nature.com/articles/d41586-026-00388-0
3•XzetaU8•49m ago•1 comments

Red teamers arrested conducting a penetration test

https://www.infosecinstitute.com/podcast/red-teamers-arrested-conducting-a-penetration-test/
1•begueradj•56m ago•0 comments

Show HN: Open-source AI powered Kubernetes IDE

https://github.com/agentkube/agentkube
2•saiyampathak•59m ago•0 comments

Show HN: Lucid – Use LLM hallucination to generate verified software specs

https://github.com/gtsbahamas/hallucination-reversing-system
2•tywells•1h ago•0 comments

AI Doesn't Write Every Framework Equally Well

https://x.com/SevenviewSteve/article/2019601506429730976
1•Osiris30•1h ago•0 comments

Aisbf – an intelligent routing proxy for OpenAI compatible clients

https://pypi.org/project/aisbf/
1•nextime•1h ago•1 comments

Let's handle 1M requests per second

https://www.youtube.com/watch?v=W4EwfEU8CGA
1•4pkjai•1h ago•0 comments

OpenClaw Partners with VirusTotal for Skill Security

https://openclaw.ai/blog/virustotal-partnership
1•zhizhenchi•1h ago•0 comments

Goal: Ship 1M Lines of Code Daily

2•feastingonslop•1h ago•0 comments

Show HN: Codex-mem, 90% fewer tokens for Codex

https://github.com/StartripAI/codex-mem
1•alfredray•1h ago•0 comments

FastLangML: FastLangML:Context‑aware lang detector for short conversational text

https://github.com/pnrajan/fastlangml
1•sachuin23•1h ago•1 comments

LineageOS 23.2

https://lineageos.org/Changelog-31/
2•pentagrama•1h ago•0 comments

Crypto Deposit Frauds

2•wwdesouza•1h ago•0 comments

Substack makes money from hosting Nazi newsletters

https://www.theguardian.com/media/2026/feb/07/revealed-how-substack-makes-money-from-hosting-nazi...
5•lostlogin•1h ago•0 comments

Framing an LLM as a safety researcher changes its language, not its judgement

https://lab.fukami.eu/LLMAAJ
1•dogacel•1h ago•0 comments

Are there anyone interested about a creator economy startup

1•Nejana•1h ago•0 comments

Show HN: Skill Lab – CLI tool for testing and quality scoring agent skills

https://github.com/8ddieHu0314/Skill-Lab
1•qu4rk5314•1h ago•0 comments
Open in hackernews

Leprechauns, root causes, and other fairy tales

https://www.tomdalling.com/blog/software-processes/leprechauns-root-causes-and-other-fairy-tails/
29•ingve•6mo ago

Comments

lucianbr•6mo ago
Air accidents investigations somehow seem to result in increased safety, so something can work. First idea that comes to mind is that those reports point out multiple causes usually.
satisfice•6mo ago
This article is itself an example of oversimplifying a complex process.

Five whys and the notion of a single root cause are both weak heuristics— but also almost no one sees them otherwise. People are generally smart, not stupid, and already understand that systems are complex.

Still, it is useful to ask “how did this problem come to be?” and “what can we do to improve our system?” Neither of which require or even encourage oversimplification.

derbOac•6mo ago
> understand that systems are complex.

For what it's worth, this is not my experience. My experience is that there's often an attempt to assign blame to a small component of the system. My sense is it is ultimately to reduce effort or embarrassment, to blame whatever can be "fixed" with less work or indictment of "higher level" components. So the littler or less well-integrated person or more minor policy gets blamed over management or administration, or core operating policies, or culture, because the latter are harder to change or involve more admission of fundamental problems that could cause reputational harm.

It's uncommon in my experience to acknowledge systemic problems or problems with very fundamental policies.

My only criticism of the piece is that sometimes you can identify a root cause, but it's at a very high level of generality and/or involves an error of omission, which is harder to identify than an error of commission. For example, a policy or protocol that isn't present that could be.

satisfice•6mo ago
"There is often an attempt..." is completely compatible with my point. SOME people oversimplify. Other people are merely simplifying. Assigning blame is also a relative matter. Some people might claim that all the blame belongs in one place, while most would admit, under cross-examination, that not ALL belongs in any one place.

Of course blame follows the lines of ideology and self-interest. This is also not indicative of people being stupid and not understanding complexity-- because it's not a problem with understanding. It's a problem of damage control.

It's entirely rational to want to control damage!

Yes, it's uncommon to acknowledge systemic problems. My point is that educating people about such things won't help much, since lack of education is not to blame for this pattern. See what I did there?

JonathanRaines•6mo ago
I think you are right that once the exercise becomes hunting for a scapegoat it's pointless.

However, it can be a way for everyone to understand the system better. The goal should be making each of the dominoes less likely to fall. Doing so can simplify rather than add complexity.