HNHacker News
TopNewBestAskShowJobs

p_l

10,056 karma · joined July 2, 2011

[ my public key: https://keybase.io/pawel_lasek; my proof: https://keybase.io/pawel_lasek/sigs/pq37gK20v7ngSJ9wYjGNZ9K7DjPDz-X-zkaE9hUuUtc ]
submissionscomments
p_l··on Private Jet plane crashes in Florida
Yep. Picked 32640 because the last two values are, from my understanding, inconsistently handled by transmitters due to how far from norm they are.
p_l··on Nvidia – open-source auto-ontology
You can self-host the LLMs involved, API key is only if you're using nvidia-hosted
p_l··on Jobless tech workers are being left out of San Francisco's AI boom
Part of the offshore story is why I have way less care for big name journalism complaining about AI job loss or the like.

I remember their cheerful atmosphere about "geeks" getting pushed back down thanks to offshoring

p_l··on Private Jet plane crashes in Florida
32,640 fpm is the maximum ADS-B can encode for vertical speed.

Quite likely it exceeded that, some sensible numbers I've seen suggest between 800 to 1000 km/h final speed at impact

p_l··on Private Jet plane crashes in Florida
This kind of impact in fact has tendency to turn aircraft into powder
p_l··on Private Jet plane crashes in Florida
It's not even cockpit audio, it's just from ATC radio : - \
p_l··on Private Jet plane crashes in Florida
The speculated final velocity at time of impact greatly exceeded design parameters for black box survivability so...
p_l··on Herdr has released native app
Same here - just started to really get going with herdr not sure I want to worry about moving again
p_l··on We're releasing the weights of Mistral Large 4
Unfortunately I can't link to IRL discussion with lawyer (wish it was so easy!)

We got guidance that:

a) weights should be considered to be public domain due to "blender" aspect of training on otherwise copyrighted data (similar argument as to why stable diffusion weights can't be infringing - because any specific copyrighted work provides less impact on process than RNG)

b) model outputs should be considered public domain as far as "work done by model" counts - prompts/inputs done by human are copyrightable.

However, model output can infringe - essentially same framework as if you used a ton of Disney stuff in a drawing you made by yourself - since the drawing would remain what it is without the copyrighted information, it is infringing even if you did it from memory.

Similarly any human-made addition to model output are copyrightable, but to make an useful distinction there would need to be some significant work. So let's say you made an image that incorporated model outputs for parts of it - going back to "prompts" it also counts if you made a sketch/wireframe/etc. and then used a model to "fill in" or postprocess it.

p_l··on We're releasing the weights of Mistral Large 4
https://www.europarl.europa.eu/RegData/etudes/BRIE/2026/7969...

Unless considerable amount of lobbying reverses it in a new directive/regulation, that's the case on model outputs - Originally I learnt of it from briefing prepared by lawyers regarding EU laws on AI.

p_l··on We're releasing the weights of Mistral Large 4
In EU especially it's already settled, afaik, as are direct model outputs
p_l··on Gemini 4 Carbon is the new #1 on the Artificial Analysis Intelligence Index v5
I guess it's also because the model is available to at least one group, and this releases embargo on information about it.
p_l··on Gemini 4 Carbon
The point of the lower pricing is to move compute (and sysadmin) capacity around
p_l··on Gemini 4 Carbon
Nah, Elon is well reported to have been crazy about "X.com" since ~2000 and thus it makes sense it got incorporated into SpaceX name
p_l··on Gemini 4 Carbon
AFAIK they are all (Barium, Argon, Carbon) different checkpoints of the same (ongoing) training run
p_l··on Code-execution flaw in UEFI firmware threatens most servers on planet Earth
Arguably it's trivial to fix for most netbooting environments, the one iffy case would be corporate laptops configured to boot from net for management purposes because there "attacker controls the network and timing" suddenly can become easy
p_l··on Code-execution flaw in UEFI firmware threatens most servers on planet Earth
"Think" (thinkpad, thinkcentre, etc.) series have AFAIK completely different R&D team plus the whole aspect of building on top of what was done for earlier models, and on those the UEFI Capsule mechanism (which allows passing firmware updates from booted OS to firmware) works just fine - it's also used to ship firmware upgrade as part of Windows Update.
p_l··on Code-execution flaw in UEFI firmware threatens most servers on planet Earth
UEFI provides a way to update in OS agnostic way, unfortunately motherboard vendors do not necessarily provide archives that can be uploaded this way (assuming they updated to new enough version to support the interface in the first place).

Unfortunately we're suffering from "eh, it boots windows, ship it" mentality of many hardware manufacturers, whether back in horrible era of BIOS emulating faux IBM PC 5150, or with modern UEFI

p_l··on Tailscale SSH: Investigation and remediation update
No, by DPI. SSH has recognizable signature as does web sockets etc.

Takes flipping one switch for 443 to be made HTTPS + HTTP2 + websocket only by egress firewall if you paid for one with DPI

p_l··on Tailscale SSH: Investigation and remediation update
Have you seen the slopreport for the vulnerability?

I fully understand being carefully worded when you need serious brain power just to figure out every claim in the logorrhea

p_l··on Show HN: Encarta 95 Web
and includes AoE 2 :D
p_l··on Tailworm – Tailscale Vulnerabilities
Reserving any judgement on the actual vulnerability or disclosure process - It was still very hard to read, and not because the topic was hard - the writing just sucked.
p_l··on Tailworm – Tailscale Vulnerabilities
The following are from my own adventures in heavily customizing headscale and interacting with tailscale client:

Nodes get "NetMap" streamed to them from control plane, which contains information about nodes accessible to them. Without a node showing up in netmap, you can't start discovery to connect to that node - discovery protocol follows once you get information for given node, and allows continued connection in case NetMap gets stale (but if ACLs are changed and reach the other node, it's going to filter you out, and you're not going to know about it without initial info from NetMap)

p_l··on Tailworm – Tailscale Vulnerabilities
Pretty sure only accessible nodes are present in netmap given to a node from headscale.

... Because I had triggered more than one alarm recently messing with a headscale upgrade and permissions involved :D

p_l··on Canadians are running out of time to oppose Bill C-22
FATCA shows how promised reciprocity works (spoiler: it does not)
p_l··on OpenSSL 4.1.0
Honestly more HPPA gear to test against than Itanium
p_l··on The same AI request emits 8.5 times more CO2 in Texas than in Sweden
Perhaps they are mistaking Sweden with Norway (whose nearly 100% hydropower system is constrained)
p_l··on A skeptic's first contact with Kubernetes (2024)
A good term for it is that every bigger system is going to be at best metastable
p_l··on A skeptic's first contact with Kubernetes (2024)
It's why, except for upstream charts, I use exclusively jsonnet (through kubecfg) when I can.

String interpolation is a pox on the ecosystem, especially given how YAML was only originally a debug helper...

p_l··on Butlerian Jihad
And which is better explained as two eugenics programs at clashing into each other, both working on dominating the general populace (Bene Gesserit arguably on larger scale than Komos' single-system thing), and Bene Gesserit using their long-worked socio-religious manipulation for revenge that gets out of hand
Page 1 of 34Next →