Arvīds Kokins
@archo5.bsky.social
Trying to make things that vaguely resemble games (and the tech to make them work). Recently released corpo/ghost as @interleavedsystems.com
https://store.steampowered.com/app/2844660/corpoghost/?utm_source=bskybio
https://store.steampowered.com/app/2844660/corpoghost/?utm_source=bskybio
"fragmentation" is another word for "more competition than we'd like to have"
the only way you can "address challenges" is by removing the competition
the only way you can "address challenges" is by removing the competition
November 19, 2025 at 9:04 PM
"fragmentation" is another word for "more competition than we'd like to have"
the only way you can "address challenges" is by removing the competition
the only way you can "address challenges" is by removing the competition
so is it an error or is it the correct form?
fucking slop lol
fucking slop lol
November 12, 2025 at 6:40 AM
so is it an error or is it the correct form?
fucking slop lol
fucking slop lol
that feeling when you finish watching a 37 minutes long video 26 minutes after it is released
October 26, 2025 at 5:18 PM
that feeling when you finish watching a 37 minutes long video 26 minutes after it is released
side note: google "AI" overview apparently does not understand English
October 24, 2025 at 7:55 PM
side note: google "AI" overview apparently does not understand English
still waiting for some food place to make a big ass button that says "EMIT CONDIMENT"
October 24, 2025 at 7:50 PM
still waiting for some food place to make a big ass button that says "EMIT CONDIMENT"
(p25) this is just.. no
- it doesn't do "analysis", "synthesis" or "summarization"
- it doesn't "discover new ideas"
it only imitates the statistically likely words generated in such circumstances
if your case matches a preexisting one well enough, you'll get an answer - otherwise, a silent error
- it doesn't do "analysis", "synthesis" or "summarization"
- it doesn't "discover new ideas"
it only imitates the statistically likely words generated in such circumstances
if your case matches a preexisting one well enough, you'll get an answer - otherwise, a silent error
![AI further contributes to productivity through the discovery of new ideas and through new efficient and
effective means of conducting work (Jones, 2022[47]). One of AI’s most notable benefits is its ability to help
government offices handle and manage the analysis and synthesis of extensive documentation
(OECD/UNESCO, 2024[12]). While a variety of AI-enabled tools may be useful, LLMs in particular can serve
as powerful assistants for government officials in this regard, aiding in tasks such as research, content
summarisation and synthesis (Berglind, Fadia and Isherwood, 2022[37]). Research conducted with
knowledge professionals in the private sector showed that AI use can improve the performance of
individual and teams, and also break down functional siloes (Dell’Acqua et al., 2025[48]). Governments can
use AI to reduce civil servants’ workload and to improve access to information for both citizens and public
servants.](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:6k47jtprfuiilruw7wlx6mlu/bafkreiffgyrr7ia6do2inlcakgtktviy5xifwmst2u6lmp7qusz66tg5la@jpeg)
October 21, 2025 at 10:31 AM
(p25) this is just.. no
- it doesn't do "analysis", "synthesis" or "summarization"
- it doesn't "discover new ideas"
it only imitates the statistically likely words generated in such circumstances
if your case matches a preexisting one well enough, you'll get an answer - otherwise, a silent error
- it doesn't do "analysis", "synthesis" or "summarization"
- it doesn't "discover new ideas"
it only imitates the statistically likely words generated in such circumstances
if your case matches a preexisting one well enough, you'll get an answer - otherwise, a silent error
www.oecd.org/content/dam/...
this report is severely undermined by not separating "AI" into specific types of tech, or even by failure mode
LLMs, OCR and pathfinding are not even close to each other
makes most of the report effectively useless
this is relatively nice however (on page 9):
this report is severely undermined by not separating "AI" into specific types of tech, or even by failure mode
LLMs, OCR and pathfinding are not even close to each other
makes most of the report effectively useless
this is relatively nice however (on page 9):
October 21, 2025 at 9:52 AM
www.oecd.org/content/dam/...
this report is severely undermined by not separating "AI" into specific types of tech, or even by failure mode
LLMs, OCR and pathfinding are not even close to each other
makes most of the report effectively useless
this is relatively nice however (on page 9):
this report is severely undermined by not separating "AI" into specific types of tech, or even by failure mode
LLMs, OCR and pathfinding are not even close to each other
makes most of the report effectively useless
this is relatively nice however (on page 9):
the answer is - more stupid than I previously thought
are they hosting everything in the US too?
this is insane
imagine effectively giving trump and random oligarchs an off-switch to your government's critical infrastructure
are they hosting everything in the US too?
this is insane
imagine effectively giving trump and random oligarchs an off-switch to your government's critical infrastructure
October 20, 2025 at 10:11 AM
the answer is - more stupid than I previously thought
are they hosting everything in the US too?
this is insane
imagine effectively giving trump and random oligarchs an off-switch to your government's critical infrastructure
are they hosting everything in the US too?
this is insane
imagine effectively giving trump and random oligarchs an off-switch to your government's critical infrastructure
www.axios.com/2025/10/07/a...
what "something productive"? the endless spam? the broken shitpost images and videos? the overpriced half-GPUs with a ~5 year lifespan or the massive buildings they're in?
holy shit these people are either fucking stupid or in damage control mode already
what "something productive"? the endless spam? the broken shitpost images and videos? the overpriced half-GPUs with a ~5 year lifespan or the massive buildings they're in?
holy shit these people are either fucking stupid or in damage control mode already
October 8, 2025 at 3:10 PM
www.axios.com/2025/10/07/a...
what "something productive"? the endless spam? the broken shitpost images and videos? the overpriced half-GPUs with a ~5 year lifespan or the massive buildings they're in?
holy shit these people are either fucking stupid or in damage control mode already
what "something productive"? the endless spam? the broken shitpost images and videos? the overpriced half-GPUs with a ~5 year lifespan or the massive buildings they're in?
holy shit these people are either fucking stupid or in damage control mode already
had a similar thought lol
September 27, 2025 at 2:48 AM
had a similar thought lol
we still effectively have gatekeepers, they're just in the media now
disagree on the "more interesting/creative", many of the games that came out in the '00s still haven't been surpassed (they're being remade/re-released now instead)
choice is a finite resource so "more choices" isn't all good
disagree on the "more interesting/creative", many of the games that came out in the '00s still haven't been surpassed (they're being remade/re-released now instead)
choice is a finite resource so "more choices" isn't all good
September 26, 2025 at 10:23 PM
we still effectively have gatekeepers, they're just in the media now
disagree on the "more interesting/creative", many of the games that came out in the '00s still haven't been surpassed (they're being remade/re-released now instead)
choice is a finite resource so "more choices" isn't all good
disagree on the "more interesting/creative", many of the games that came out in the '00s still haven't been surpassed (they're being remade/re-released now instead)
choice is a finite resource so "more choices" isn't all good
is that an "AI" summary? 👀
September 24, 2025 at 7:16 PM
is that an "AI" summary? 👀
the "far left": oh no not our dear *billionaires*!
x.com/Cernovich/st...
I can't even tell if this is parody
x.com/Cernovich/st...
I can't even tell if this is parody
September 10, 2025 at 8:50 PM
the "far left": oh no not our dear *billionaires*!
x.com/Cernovich/st...
I can't even tell if this is parody
x.com/Cernovich/st...
I can't even tell if this is parody
looks like that A100 GPU was sold for ~$2000 👀
September 7, 2025 at 3:22 PM
looks like that A100 GPU was sold for ~$2000 👀
the clock is upside down, it's actually 🪑🪑:🪑 chair time
August 11, 2025 at 3:07 AM
the clock is upside down, it's actually 🪑🪑:🪑 chair time
www.menomonieminute.com/post/city-pr...
"Balloonist LLC" lol
those data center bozos can't possibly be serious
"Balloonist LLC" lol
those data center bozos can't possibly be serious
August 6, 2025 at 3:41 AM
www.menomonieminute.com/post/city-pr...
"Balloonist LLC" lol
those data center bozos can't possibly be serious
"Balloonist LLC" lol
those data center bozos can't possibly be serious
2 concerns:
- if many do decide to downscope, that would still lead to lots of layoffs, wouldn't it?
- as for years of preproduction - that stage is supposed to be very cheap and use a small fraction of the workforce even if it takes years (would expect it to take a few % of the total dev budget)
- if many do decide to downscope, that would still lead to lots of layoffs, wouldn't it?
- as for years of preproduction - that stage is supposed to be very cheap and use a small fraction of the workforce even if it takes years (would expect it to take a few % of the total dev budget)
August 1, 2025 at 7:24 PM
2 concerns:
- if many do decide to downscope, that would still lead to lots of layoffs, wouldn't it?
- as for years of preproduction - that stage is supposed to be very cheap and use a small fraction of the workforce even if it takes years (would expect it to take a few % of the total dev budget)
- if many do decide to downscope, that would still lead to lots of layoffs, wouldn't it?
- as for years of preproduction - that stage is supposed to be very cheap and use a small fraction of the workforce even if it takes years (would expect it to take a few % of the total dev budget)
I don't have the nice test scene on my PC but here's something
(2 slices, the nearest one is 90deg FOV and the other one is 45)
it indeed looks very similar to regular perspective but it feels strange when moving or rotating the camera
(2 slices, the nearest one is 90deg FOV and the other one is 45)
it indeed looks very similar to regular perspective but it feels strange when moving or rotating the camera
July 27, 2025 at 8:26 PM
I don't have the nice test scene on my PC but here's something
(2 slices, the nearest one is 90deg FOV and the other one is 45)
it indeed looks very similar to regular perspective but it feels strange when moving or rotating the camera
(2 slices, the nearest one is 90deg FOV and the other one is 45)
it indeed looks very similar to regular perspective but it feels strange when moving or rotating the camera
July 24, 2025 at 4:54 PM
as for "problems it hasn’t seen before", the search results clearly indicate 61 web pages, among those - another Verse paper
any search results require text extraction and indexing
can't say if the paper was also used to build the LLM (or for RAG) but it's certainly possible
any search results require text extraction and indexing
can't say if the paper was also used to build the LLM (or for RAG) but it's certainly possible
![Search Results
[PDF] a Core Calculus for Functional Logic Programming
In this paper we describe the Verse calculus, VC, a new core calculus for functional logical programming. Our main contribution is to equip VC with a small-step ...
simon.peytonjones.org](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:6k47jtprfuiilruw7wlx6mlu/bafkreidlkiklfem3zalo67m6qtwea2zqlniloyzz345dz3gzky3mveqbee@jpeg)
July 16, 2025 at 2:28 PM
as for "problems it hasn’t seen before", the search results clearly indicate 61 web pages, among those - another Verse paper
any search results require text extraction and indexing
can't say if the paper was also used to build the LLM (or for RAG) but it's certainly possible
any search results require text extraction and indexing
can't say if the paper was also used to build the LLM (or for RAG) but it's certainly possible
linked example: grok.com/share/bGVnYW...
even a purely superficial reading of it finds issues:
"without succumbing to non-determinism" contains a double negative and redundant words and can be rewritten without both of those
who finds linguistic spam useful or helpful when trying to learn things?
even a purely superficial reading of it finds issues:
"without succumbing to non-determinism" contains a double negative and redundant words and can be rewritten without both of those
who finds linguistic spam useful or helpful when trying to learn things?
July 16, 2025 at 1:57 PM
linked example: grok.com/share/bGVnYW...
even a purely superficial reading of it finds issues:
"without succumbing to non-determinism" contains a double negative and redundant words and can be rewritten without both of those
who finds linguistic spam useful or helpful when trying to learn things?
even a purely superficial reading of it finds issues:
"without succumbing to non-determinism" contains a double negative and redundant words and can be rewritten without both of those
who finds linguistic spam useful or helpful when trying to learn things?
it said "new technologies or services", not meaningfully limiting the scope in any way
moderation ML goes after "including", so it's just one example, not a complete list
moderation ML goes after "including", so it's just one example, not a complete list
![You hereby grant us a perpetual, worldwide, non-exclusive, royalty-free, transferable, sub-licensable license to use your Content for the purposes of operating, developing, commercializing, and improving the Service or new technologies or services, including [...]](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:6k47jtprfuiilruw7wlx6mlu/bafkreidxdpql7m5dcevoz2jyvlsus2ue42hgalv2yhxfiktsqxr3vbray4@jpeg)
July 15, 2025 at 2:04 PM
it said "new technologies or services", not meaningfully limiting the scope in any way
moderation ML goes after "including", so it's just one example, not a complete list
moderation ML goes after "including", so it's just one example, not a complete list