ec109685

35 ms·

601.
by ec109685 3y ago
This is an interesting benchmark that compares disk read latency across different clouds across various disk configurations, including SSD's and EBS: https://github.com/scylladb/diskplorer Google's disk perfo
602.
by ec109685 3y ago
Can’t that be amortized across all videos (e.g. if YouTube had a decompressor they downloaded once)?
603.
by ec109685 3y ago
The example from the article was “pet air purifier”, with fortune.com leading the ranking.
604.
by ec109685 3y ago
GenAI as in increasingly sophisticated reasoning engines, not writing fake product reviews.
605.
by ec109685 3y ago
A site that is excellent at SEO spam isn’t the same as a “democratically elected” result. The list of sites for that term are utter crap, recommending the purifiers paying the top commission. Why would you want Google to perpetuate that ran
606.
by ec109685 3y ago
It’s acknowledged obliquely at the end of this tweet: https://twitter.com/searchliaison/status/1716964371916800472
607.
by ec109685 3y ago
This type of SEO spam seems like an area where generative AI (i.e. sophisticated reading systems) will be able to address. If there was a human curator with infinite time, I think they could wade through this crap and find ten links to auth
608.
by ec109685 3y ago
The article points out numerous examples where a site like people.com is ranking for pet air purifiers, with zero evidence that they actually tested the products in question. This tweet thread goes into more detail https://x.com&
609.
by ec109685 3y ago
Are there services that have a copy of the Internet for a fee? How did perplexity.ai crawl the internet for their AI?
610.
by ec109685 3y ago
Amazon selling obviously fake SD cards / flash disks is baffling. It’s super easy to spot and fake disks cost users data, so you’d think they’d take that seriously. https://a.co/d/auXcxrZ ($29 for 1TB as their “ov
611.
by ec109685 3y ago
This is impressive, but other parts of Windows are so dreary. Installs of apps that throw up all over the disk, Windows Updates that mysteriously fail in unrecoverable ways 87% of the way through and cryptic error codes and procedures to di
612.
by ec109685 3y ago
There’s three options for inference: 1) On device inference 2) Inference “on the edge” 3) Inference in a data center Given fly is deployed in equinox data centers just like everyone else, fundamentally there isn’t much difference betw
613.
by ec109685 3y ago
Tesla doesn’t let you alter the car’s course without enough force to disengage.
614.
by ec109685 3y ago
The recipe example or any any LLM use case seems like a very poor way of highlighting “inference at the edge” given the extra few hundred ms round trip won’t matter.
615.
by ec109685 3y ago
Teslas don’t require much force at all to disengage.
616.
by ec109685 3y ago
This housing is deed restricted to certain income levels.
617.
by ec109685 3y ago
Often some of the most impressive feats are just parlor tricks. E.g. Gemini might have been trained on data like this: https://www.google.com/search?q=Spell+the+word+lollipop+back...
618.
by ec109685 3y ago
Apple was paid a monthly fee by Cigular per user for the iPhone 1, so it was just subsidized weirdly.
619.
by ec109685 3y ago
If you used copilot in the beginning (and I think still with some plans), it was only GPT 3.5. Likely you’d get much better results with GPT-4.
620.
by ec109685 3y ago
There’s a great answer here that describes a rolling code attack and above it, an answer describing that they have slack regarding where they are in the code sequence. https://crypto.stackexchange.com/a/47440
621.
by ec109685 3y ago
If you have control of the opener, couldn’t you use the door’s learning mode and make it into a real opener?
622.
by ec109685 3y ago
It’s still trunk based development even if a feature branch / tag is created that will match production.
623.
by ec109685 3y ago
How did the hold back end up working out? Was the learning worth the hassle?
624.
by ec109685 3y ago
YouTube Premium exists to remove ads.
625.
by ec109685 3y ago
This link lets individuals opt out: https://privacy.openai.com/policies
626.
by ec109685 3y ago
I was responding to this part, “And I’d venture their standards match their users’ better than the author’s”, which I understood you to mean people like these seo’d results.
627.
by ec109685 3y ago
The deck: https://media.defcon.org/DEF%20CON%2031/DEF%20CON%2031%20pre... Quite reminiscent of the exploits the JVM suffered with reflection and deserialization. Fascinating how many iterations it took Apple to mitigat
628.
by ec109685 3y ago
https://getthatvideo.com/ Is the first result for downloading YouTube videos. Seems super sus (especially since the site doesn’t load). Auto-prompted to: "Here's a helpful website for downloading YouTube videos:&q
629.
by ec109685 3y ago
Kagi works because there is no incentive for SEO manipulators to target it since their market share is so small. Super tinfoil hat to believe Google wants to send users to blog spam websites (e.g. beneficial to Google). Anytime there is mon
630.
by ec109685 3y ago
Why would an average user want blog spam search results? My hope is as LLM’s improve, they can be more discriminating about the results returned.

More ›