Logs: liberachat/#haskell
| 2025-11-26 18:57:01 | <haskellbridge> | <sm> * code, which prompting strategy |
| 2025-11-26 18:57:23 | <haskellbridge> | <sm> oops irc, I shouldn't edit |
| 2025-11-26 18:57:29 | <Square3> | But yeah, I get that it's highly dependent on task / tools. Here I was mostly curious on the general view of the concept. |
| 2025-11-26 18:58:23 | <EvanR> | "an authority" is also highly subjective... I was just caught up on this 4 year old drama involving a minecraft speed runner with a hoard of fans defending them at all costs |
| 2025-11-26 18:58:36 | <EvanR> | after they cheated |
| 2025-11-26 18:58:44 | <EvanR> | so pick your authorities carefully |
| 2025-11-26 18:59:07 | <EvanR> | you could imagine someone like this shilling AI code that is just awful |
| 2025-11-26 18:59:24 | <haskellbridge> | <sm> I think it's pretty much settled by now that AI can, is, and will be increasingly helpful in programming generally speaking |
| 2025-11-26 18:59:46 | <EvanR> | as long as it still exists in the current form |
| 2025-11-26 19:00:02 | <EvanR> | which has high negative profits |
| 2025-11-26 19:00:09 | <EvanR> | that can't go on |
| 2025-11-26 19:04:58 | <haskellbridge> | <sm> 👍️ yes I mean the technology when its available |
| 2025-11-26 19:05:57 | <EvanR> | the soliton radar is made from currently existing technology |
| 2025-11-26 19:13:22 | <Leary> | Imo, to make good use of AI in programming we just need an iterative approach with both the human and the compiler in the loop. Instead of trying to "engineer" prompts for LLMs trained to replicate all text in existence, we use smaller, specialised machines trained on the much richer semantic data produced by the compiler. One for type errors, one filling holes from context, one fixing bugs given a test failure, etc. Initial data can be gathered by putt |
| 2025-11-26 19:13:23 | <Leary> | ing random errors and holes in existing programs, and more collected every time a programmer accepts/rejects an output... |
| 2025-11-26 19:14:21 | <EvanR> | it sounds like that has a chance of running locally |
| 2025-11-26 19:15:46 | → | Googulator89 joins (~Googulato@2a01-036d-0106-4ad8-f42e-6d50-f4ab-2863.pool6.digikabel.hu) |
| 2025-11-26 19:15:56 | × | Googulator45 quits (~Googulato@2a01-036d-0106-4ad8-f42e-6d50-f4ab-2863.pool6.digikabel.hu) (Quit: Client closed) |
| 2025-11-26 19:17:19 | × | sindu quits (~sindu@2.148.32.207.tmi.telenormobil.no) (Ping timeout: 240 seconds) |
| 2025-11-26 19:30:05 | → | tromp joins (~textual@2001:1c00:3487:1b00:c5b7:b8d9:7db7:74e1) |
| 2025-11-26 19:32:58 | × | Googulator89 quits (~Googulato@2a01-036d-0106-4ad8-f42e-6d50-f4ab-2863.pool6.digikabel.hu) (Quit: Client closed) |
| 2025-11-26 19:33:05 | → | peterbecich joins (~Thunderbi@172.222.148.214) |
| 2025-11-26 19:33:16 | → | Googulator89 joins (~Googulato@84-236-53-137.pool.digikabel.hu) |
| 2025-11-26 19:35:54 | <monochrom> | About current AI ("AI") and the training data: Those LLMs that have learned from history are doomed to repeat it. >:) |
| 2025-11-26 19:36:04 | × | EvanR quits (~EvanR@user/evanr) (Remote host closed the connection) |
| 2025-11-26 19:36:24 | → | EvanR joins (~EvanR@user/evanr) |
| 2025-11-26 19:37:17 | <monochrom> | Leary: That sounds dangerously close to the water fall model that every hipster agile coder has denounced. |
| 2025-11-26 19:37:54 | <monochrom> | (Don't worry, I denounce back those hipster agile fads.) |
| 2025-11-26 19:38:27 | <int-e> | hey picture moment: first you sell them agile coding, then you sell the crutches (LLMs) |
| 2025-11-26 19:39:01 | <int-e> | (and inbetween people break a few legs) |
| 2025-11-26 19:39:58 | → | sindu joins (~sindu@2.148.32.207.tmi.telenormobil.no) |
| 2025-11-26 19:40:53 | <monochrom> | Hey speaking of "synergy" business tactics. I saw this beautiful scene at a mall: A gym and a dessert pastry bakery next to each other. |
| 2025-11-26 19:48:55 | × | peterbecich quits (~Thunderbi@172.222.148.214) (Ping timeout: 240 seconds) |
| 2025-11-26 19:51:19 | <EvanR> | based |
| 2025-11-26 19:51:34 | → | statusbot joins (~statusbot@ec2-34-198-122-184.compute-1.amazonaws.com) |
| 2025-11-26 19:51:50 | × | statusbot3 quits (~statusbot@ec2-34-198-122-184.compute-1.amazonaws.com) (Remote host closed the connection) |
| 2025-11-26 19:53:10 | → | pavonia joins (~user@user/siracusa) |
| 2025-11-26 20:03:04 | × | Shark8 quits (~Shark8@c-174-56-102-109.hsd1.nm.comcast.net) (Read error: Connection reset by peer) |
| 2025-11-26 20:03:15 | → | Shark8 joins (~Shark8@c-174-56-102-109.hsd1.nm.comcast.net) |
| 2025-11-26 20:08:16 | × | ouilemur quits (~jgmerritt@user/ouilemur) (Quit: WeeChat 4.7.0) |
| 2025-11-26 20:15:59 | → | ouilemur joins (~jgmerritt@user/ouilemur) |
| 2025-11-26 20:20:40 | → | merijn joins (~merijn@77.242.116.146) |
| 2025-11-26 20:22:06 | × | jreicher quits (~user@user/jreicher) (Quit: In transit) |
| 2025-11-26 20:26:23 | → | driib3180 joins (~driib@vmi931078.contaboserver.net) |
| 2025-11-26 20:27:03 | Googulator89 | is now known as Googulator |
| 2025-11-26 20:27:20 | × | Anarchos quits (~Anarchos@91-161-254-16.subs.proxad.net) (Quit: Vision[]: i've been blurred!) |
| 2025-11-26 20:37:20 | × | trickard_ quits (~trickard@cpe-86-98-47-163.wireline.com.au) (Read error: Connection reset by peer) |
| 2025-11-26 20:37:33 | → | trickard_ joins (~trickard@cpe-86-98-47-163.wireline.com.au) |
| 2025-11-26 20:49:08 | → | Frostillicus joins (~Frostilli@pool-71-174-119-69.bstnma.fios.verizon.net) |
| 2025-11-26 20:50:56 | × | yin quits (~zero@user/zero) (Ping timeout: 240 seconds) |
| 2025-11-26 20:51:26 | → | peterbecich joins (~Thunderbi@172.222.148.214) |
| 2025-11-26 20:59:02 | trickard_ | is now known as trickard |
| 2025-11-26 21:00:04 | → | Sgeo joins (~Sgeo@user/sgeo) |
| 2025-11-26 21:01:34 | → | yin joins (~zero@user/zero) |
| 2025-11-26 21:01:50 | × | merijn quits (~merijn@77.242.116.146) (Ping timeout: 244 seconds) |
| 2025-11-26 21:15:15 | × | peterbecich quits (~Thunderbi@172.222.148.214) (Ping timeout: 265 seconds) |
| 2025-11-26 21:34:11 | → | hseg_ joins (~hseg_@46.120.21.70) |
| 2025-11-26 21:35:07 | → | jreicher joins (~user@user/jreicher) |
| 2025-11-26 21:41:06 | × | dyniec quits (~dyniec@dybiec.info) (Remote host closed the connection) |
| 2025-11-26 21:41:43 | → | dyniec joins (~dyniec@dybiec.info) |
| 2025-11-26 21:49:55 | × | Frostillicus quits (~Frostilli@pool-71-174-119-69.bstnma.fios.verizon.net) (Ping timeout: 240 seconds) |
| 2025-11-26 21:56:16 | × | takuan quits (~takuan@d8D86B9E9.access.telenet.be) (Remote host closed the connection) |
| 2025-11-26 21:57:45 | <jreicher> | I suspect AI is more useful in languages that require a fair amount of boilerplate to get things done. |
| 2025-11-26 22:00:06 | <haskellbridge> | <sm> that's us.. |
| 2025-11-26 22:00:23 | × | target_i quits (~target_i@user/target-i/x-6023099) (Quit: leaving) |
| 2025-11-26 22:00:25 | → | Frostillicus joins (~Frostilli@pool-71-174-119-69.bstnma.fios.verizon.net) |
| 2025-11-26 22:04:37 | → | peterbecich joins (~Thunderbi@172.222.148.214) |
| 2025-11-26 22:18:20 | <jackdk> | jreicher: I have seen three arguments and I'm not sure which to weight most heavily: 1. it works best on things it's seen the most of in the training distribution (python, TS, specific major libraries — Anthropic called this out in its article on the design and implementation of Claude Code); 2. it works best on strongly-typed languages because it can converge on a solution with compiler assistance (Terry Tao's posts about Lean may apply); ... |
| 2025-11-26 22:19:21 | <jackdk> | ... 3. it works best on languages with simpler syntax and less compiler smarts because the token stream just carries more information (a Gleam advocate mentioned this to me once). |
| 2025-11-26 22:21:30 | × | michalz quits (~michalz@185.246.207.193) (Remote host closed the connection) |
| 2025-11-26 22:22:49 | × | Frostillicus quits (~Frostilli@pool-71-174-119-69.bstnma.fios.verizon.net) (Ping timeout: 260 seconds) |
| 2025-11-26 22:23:24 | <geekosaur> | I lean strongly toward #1 because the only way I can see for an LLM to have boosted "AI" is for it to be a smarter Markov bot |
| 2025-11-26 22:23:55 | <jreicher> | jackdk: FWIW I "imagine" those points are very plausible, but I meant something more specific than "works". I mean USEFUL. If the AI succeeds in writing code that would have been no more effort and no slower for a human to write, it's not useful. And very concise and obvious code is like this. Lengthy boilerplate is where the AI has the potential to save time. |
| 2025-11-26 22:24:44 | → | Frostillicus joins (~Frostilli@pool-71-174-119-69.bstnma.fios.verizon.net) |
| 2025-11-26 22:28:10 | <geekosaur> | this is also flavored by my monitoring various kinds of science news, and an active area of research is hybridizing LLMs with other varieties of AI more capable of some form of reasoning (fsvo) about the data they've been trained with |
| 2025-11-26 22:29:07 | × | peterbecich quits (~Thunderbi@172.222.148.214) (Ping timeout: 264 seconds) |
| 2025-11-26 22:34:37 | × | divlamir quits (~divlamir@user/divlamir) (Read error: Connection reset by peer) |
| 2025-11-26 22:35:03 | → | divlamir joins (~divlamir@user/divlamir) |
| 2025-11-26 22:39:56 | × | gmg quits (~user@user/gehmehgeh) (Quit: Leaving) |
| 2025-11-26 22:40:13 | × | tromp quits (~textual@2001:1c00:3487:1b00:c5b7:b8d9:7db7:74e1) (Quit: My iMac has gone to sleep. ZZZzzz…) |
| 2025-11-26 22:43:55 | → | merijn joins (~merijn@host-vr.cgnat-g.v4.dfn.nl) |
| 2025-11-26 22:50:00 | × | califax quits (~califax@user/califx) (Quit: ZNC 1.8.2 - https://znc.in) |
| 2025-11-26 22:50:21 | → | califax joins (~califax@user/califx) |
| 2025-11-26 22:50:31 | × | merijn quits (~merijn@host-vr.cgnat-g.v4.dfn.nl) (Ping timeout: 240 seconds) |
| 2025-11-26 22:54:49 | → | emmanuelux joins (~emmanuelu@user/emmanuelux) |
| 2025-11-26 22:56:48 | <haskellbridge> | <loonycyborg> "AI" is really confusing term |
| 2025-11-26 22:56:54 | <haskellbridge> | <loonycyborg> I wish they just stick with LLM |
| 2025-11-26 22:57:11 | <haskellbridge> | <loonycyborg> because AI is also automated players in computer games |
| 2025-11-26 22:57:15 | × | emmanuelux quits (~emmanuelu@user/emmanuelux) (Remote host closed the connection) |
| 2025-11-26 22:57:18 | <haskellbridge> | <loonycyborg> and lots of other things |
| 2025-11-26 22:58:18 | <jreicher> | I reckon you could argue that a language server is a form of AI. :) |
| 2025-11-26 22:58:40 | <haskellbridge> | <loonycyborg> so is a buildsystem |
| 2025-11-26 22:58:44 | <haskellbridge> | <loonycyborg> because it's a form of expert system |
| 2025-11-26 22:58:48 | → | merijn joins (~merijn@host-vr.cgnat-g.v4.dfn.nl) |
| 2025-11-26 22:59:44 | × | ThePenguin quits (~ThePengui@cust-95-80-28-221.csbnet.se) (Ping timeout: 244 seconds) |
| 2025-11-26 22:59:44 | → | gmg joins (~user@user/gehmehgeh) |
| 2025-11-26 23:00:41 | → | ThePenguin joins (~ThePengui@cust-95-80-28-221.csbnet.se) |
| 2025-11-26 23:03:28 | × | merijn quits (~merijn@host-vr.cgnat-g.v4.dfn.nl) (Ping timeout: 256 seconds) |
| 2025-11-26 23:07:26 | <jreicher> | Well I guess at a high level it's all the same thing in the end. If the "smart" system is not an LLM it just means the training data has been fed to the human (in the form of experience and/or user feedback) and the human translates that into the rules (weights) of the system. |
| 2025-11-26 23:09:21 | × | __monty__ quits (~toonn@user/toonn) (Quit: leaving) |
All times are in UTC.