Posts by dvogel@mastodon.social
(DIR) Post #AS8oqwLH4Td4YmCB84 by dvogel@mastodon.social
2023-01-29T16:46:26Z
0 likes, 0 repeats
@simon you'd likely enjoy this episode of the Ezra Klein podcast. There's an excellent tangent re: "bullshit" as it pertains to ChatGPT et al.https://www.nytimes.com/2023/01/06/opinion/ezra-klein-podcast-gary-marcus.html
(DIR) Post #ASBL8Uu1PnkXWGmfZY by dvogel@mastodon.social
2023-01-30T21:37:23Z
0 likes, 0 repeats
@simon is this not what @import is for? Write out the styling in one stylesheet and then @import it within the context of each selector. I haven't done CSS stuff in a while but I feel like this worked across modern browsers as of a couple years ago.
(DIR) Post #ASHCTRehwjHTu79OVs by dvogel@mastodon.social
2023-02-02T17:48:58Z
0 likes, 0 repeats
@simon this reminds me of a great talk by @bcantrill: https://m.youtube.com/watch?v=4PaWFYm0kEwWhen I left Crowd Compass I was the last in a long string of longer-tenured engineers to leave. I was concerned about this very thing so I spent my last two days writing down the entire history of the development of the product as it had been told to me when I started, up and through my time there. I've often wondered if/how that material was ever useful to anyone left maintaining the product.
(DIR) Post #AUNfD4iTEGlRk4EM9g by dvogel@mastodon.social
2023-04-06T16:11:43Z
0 likes, 0 repeats
@simon Since we're discussing a potential change in mode without changing the goal of having reliable information, the degree of difference in quality is what matters. The difference between the NYT and ChatGPT is the same difference between the NYT and the The Register. Notable here because many fewer people would describe The Register as having a goal of providing reliable information, even though they rely on the same editorial mode as NYT.
(DIR) Post #AVYD8wz4PlddS972rA by dvogel@mastodon.social
2023-05-11T16:23:37Z
0 likes, 0 repeats
@simon this is because prompt injections are no different from other forms of user input. You cannot trust it. The mistake is in the initial conception. You cannot use LLMs under user control to replace labor for anything other than helping the user, thereby aligning their interests with yours and making attacks counter-productive. Allowing unverified LLM output to influence business processes will never be secure in the same way that allowing a user to edit your CRM database would be insecure.
(DIR) Post #AVqvIop9NXJI7KiDFg by dvogel@mastodon.social
2023-05-13T05:28:49Z
0 likes, 1 repeats
Fred sure can be a goofball sometimes! #Caturday
(DIR) Post #AXST0TlqJq9oaYUvE8 by dvogel@mastodon.social
2023-07-07T17:16:35Z
1 likes, 0 repeats
@ebassi @brennen That is perhaps true of polls run directly by an open source project. Those projects could pay for the expertise required to run polls effectively and analyze the resulting data holistically.Telemetry is also statistically biased. There is zero assurance that you're assessing the entire population of users. You're effectively oversampling users for whom the software already works well. You cannot appropriately develop weights to counter that because of the population problem.
(DIR) Post #AZpYrcYvsDF1JXH2w4 by dvogel@mastodon.social
2023-09-16T16:46:43Z
0 likes, 0 repeats
@pixel A lot of these companies are decent companies but they aren't really VC style 10-100x investments as going concerns. I think the VC angle is essentially building out services enough to sell them to Google or Microsoft as additions to their cloud services to compete with AWS. Users clearly was to be put as technologically close to a solution as possible but it is difficult for Google to reinvest in GAE given that GAE already is serviceable. They bigger bets to make with their cash for now.
(DIR) Post #AaKa0OSlbgBtHv4vhY by dvogel@mastodon.social
2023-10-01T15:51:09Z
0 likes, 0 repeats
@simon The comparisons to crypto are not based on a lack of uses. They arr based on a mismatch between promises being made and the actual potential LLMs have. The similarity is in the resulting over-investment and inevitable bust. The actual uses for LLMs are manyfold over crypto but the investment is also manyfold over crypto, adjusting for timeline.I've seen very little over-promising from you personally though. So when people criticize LLMs I don't think they are applicable to your work.
(DIR) Post #Aas8uSHxiZFJuLqlM0 by dvogel@mastodon.social
2023-10-17T20:31:45Z
0 likes, 0 repeats
Fred couldn't wait until #Caturday. He demanded a portrait be taken immediately.#CatsOfMastodon
(DIR) Post #AbCCFJ9tuGj5kmfiu8 by dvogel@mastodon.social
2023-10-27T12:40:39Z
0 likes, 0 repeats
@mjg59 My gut feeling is that is either a 180 page book without code or a 650 page book with code. A publisher is going to turn it into a 300 page book with just enough code to distract but not enough to be useful.
(DIR) Post #AbgD6pUChxJJK1qhfs by dvogel@mastodon.social
2023-11-11T00:10:51Z
0 likes, 1 repeats
My employer (Ad Hoc, we're hiring!) took @simon up on this idea a while back. It was fantastic! Dozens of our engineers gained exposure to Datasette and SQLite. Simon let us help shape the talk ahead of time and he was astoundingly capable of shifting the talk on the fly. Afterward my team explored using this setup with Litestream for replication as an alternative to AWS's Athena. Speaking with Simon saved us probably 8 hours of exploratory work.https://fedi.simonwillison.net/@simon/111388844510128715https://adhoc.team/join/jobs/
(DIR) Post #Ac6j2zGZuMtgmA6JLk by dvogel@mastodon.social
2023-11-23T19:13:33Z
0 likes, 0 repeats
@simon I don't know of any existing libraries but it should be fairly straight forward to use the ast stdlib module to parse the input and then eval it only after ensuring the tree only contains your limited vocabulary.
(DIR) Post #AcqBdMqFHV8xEjfrNI by dvogel@mastodon.social
2023-12-15T17:30:35Z
0 likes, 0 repeats
@simon @matt @danilo I get where you're going with this and I don't disagree, but in the spirit of sharpening this argument, it needs to be paired with additional points. On it's own it sounds like you're saying we should hold ourselves back from calling out horoscopes as cleverly worded nonsense that have no connection to reality because for many people the content seems to line up quite well with their lived experience of reading their horoscope daily.
(DIR) Post #AcqBdP9Sh1heOugEbY by dvogel@mastodon.social
2023-12-15T17:32:43Z
0 likes, 0 repeats
@simon @matt @danilo It might help to reframe this around a dividing point between the useful parts of LLM tech and the distracting parts. As people hone in on specific useful types of LLM outputs their lived experience will include fewer of the distractions, and thus the critique will lose credibility. This separates it from something like a horoscope which only provide distractions.
(DIR) Post #ArQtD2rFqW8oSCBgSe by dvogel@mastodon.social
2025-02-24T03:28:36Z
0 likes, 0 repeats
@andrewrk It has been 0 days since a GitHub bug for me. Just yesterday I was half way writing up an issue and their fancy shmancy new React editor locked up, refusing to moving the edit cursor at all.
(DIR) Post #AvTJozSsAokt4nLrxw by dvogel@mastodon.social
2025-06-22T19:59:40Z
0 likes, 0 repeats
@lproven @davidgerard I have no love for Weigelt. Quite the opposite. But wow you could not design a better way to kill X than to make Weigelt see himself as a martyr, making news about a fork. Now everyone who prefers to use X over Wayland has to clarify which X distro they associate with. They have reified the Wayland-esque balkanization in X land.