I should have added a smirk to my response π
I should have added a smirk to my response π
But to be fair, he had a point.
OpenBSD is having a #j2k25 hackathon in Japan. Kinda wish I was there, good times.
Iβm not against AI but the fuzz about replacing all devs is mostly a myth. Primarily to sell AI tools, to pump their company value, or to drop the comp rates of software engineers who still need to do the work - or to cleanup the mess.
Very often LLM generated code doesnβt even work from the beginning unless youβre solving very trivial problems. And you cannot expect anything above average if you use models that are trained by the big βaverageβ quality out there. And thatβs true for even the latest models.
So CEOs like Nadella say that 30% and soon most of their code is written by AI. Itβs a sales pitch or they admit having lots of bad code. I run dev teams and we have all the AI tools, but weβre very far from trusting them. Itβs not about tweaking Cursor rules, itβs about solving problems cleanly.
I get the joke, but canβt there be articles that are critical of some aspects of AI but not all of it?
I wonder if people in cyber security already see a noticeable spike of incidents due to all the AI generated code out there.
You can build great agents in languages like Rust, but Python goes deeper into ML inference, the bridge between models and apps. Formats like ONNX exist for other languages but remain unpopular as most AI devs only know Python. So weβre stuck with a scripting language in production use cases.
Itβs a familiar pattern: AI/ML was built in Python during the research phase and now most libraries and API examples are in Python as well. As a result, AI devs stick to Python, reinforcing its use. This is pushing us back by several years when low latency and distribution werenβt that critical.
Stefan Sperling (stsp@) recently enabled ice(4), a driver for Intel E810 Ethernet devices in #OpenBSD -current, and is still looking for testers who have the hardware to test on.
Stefan has been porting it from FreeBSD the over past few months requested by Genua, with funding.
CAPTCHA
Thereβs an influx of new users following me. I havenβt posted anything in over a year. Are you robots?
Itβs a bit weird to post my first OpenBSD diff after 3 years. Nothing special, just a PCI Id. https://marc.info/?l=openbsd-tech&m=168785615802294&w=2
I was serious. Welcome!
I just deleted some domains that I never used. Thatβs a first.
alright folks, the app code is now public
https://github.com/bluesky-social/social-app
Welcome! Everything is fine.
Whatβs that, OpenBSDβs puffy in a protective cocoon? Maybe some kind of VM running inside OpenBSDβs vmm?
Weird Midjourney βOpenBSD blowfishβ creations.
Trapped in a vi editor instance, commands to exit echo uselessly. Time blurs as the maze of code lines and command modes become a relentless prison. The once familiar hum of the hard drive intensifies, as hope of escape from this digital Backroom fades into oblivion.
robot cat
Iβm more of a dog person, but hereβs a cat picture by popular demand.
We'll soon be launching a sandbox environment to begin the testing phase of federation for the AT Protocol with allow-listed servers.
In advance of this launch, here are some technical details about our design decisions.
https://blueskyweb.xyz/blog/5-5-2023-federation-architecture
We need more BSD people in BlueSky, I want to see more console screenshots, puffy stickers, and panic traces in the βWhatβs hotβ stream.
Huge news! Starting today, you can set up a passkey for your Google account. At this stage of the industry-wide transition, setting up a passkey doesnβt invalidate your password, so itβs 100% safe to set up. https://blog.google/technology/safety-security/the-beginning-of-the-end-of-the-password/
@chad.ch good to see you here!
I think thatβs why many platforms support passkeys but not the password-less logins. Adding passkeys as 2FA is easy once you support WebAuthn, but skipping the password step can be a bigger change on the server side with some side-effects (eg. unauthenticated challenge retrieval).
Depends on the MFA: passkeys use the FIDO2 WebAuthn web protocol that is used for 2FA tokens like Yubikeys. Once you support it, you already support passkeys. Theyβre supposed to be a password replacement, so the platform has to skip the password in the login flow β this is technically not required.