Austin Wallace

5K posts

Austin Wallace banner
Austin Wallace

Austin Wallace

@austeane

System optimizer: Data engineering, AI tooling, and Team Canada quadball

Victoria, British Columbia Katılım Aralık 2009
1.3K Takip Edilen326 Takipçiler
Austin Wallace
Austin Wallace@austeane·
@theo It works well, you don't even have to overthink it. Plan mode, make sure the plan is testable Tell it to complete the feature and test it, including using claude in chrome. It isn't perfect but is way better than what you got out of it! I use probably 75% Codex 25% Opus
English
0
0
0
31
Theo - t3.gg
Theo - t3.gg@theo·
Just let Opus go for over an hour on a new feature. When it was done, I asked how I can test it. 20 minutes later, it realized I can't test it because it did the whole thing entirely wrong. Idk how you guys use this model every day for real work 🙃
Theo - t3.gg tweet media
English
373
24
1.4K
234.1K
Austin Wallace
Austin Wallace@austeane·
@AndreaCris74423 @LotharHS What they left intact was really only his ability to entry and take space one time. Every creative way to play him was majorly nerfed, and his entry in ranked is way worse than multiple other duelists.
English
0
0
0
12
andy_vlr
andy_vlr@AndreaCris74423·
@austeane @LotharHS Removing one flash just completely killed him imo, I would have preferred removing 1 tp and maybe reducing the ult timer from 10 to 8 seconds, this just made him so shit from every point of view
English
1
0
0
21
Austin Wallace
Austin Wallace@austeane·
@LotharHS It is trending that way. I really wish they could release/support an agent that is is about that creativity and has a high skill ceiling, though I understand that that agent shouldn’t *also* be an entry flash agent
English
0
0
1
132
DRX LotharHS
DRX LotharHS@LotharHS·
@austeane The game doesn’t reward putting effort and creativity anymore. It’s unfortunate but the devs showed which kind of community they want to appeal to, by making consecutive changes over the course years now.
English
1
0
10
564
Austin Wallace
Austin Wallace@austeane·
@Sathos__voice @jachiam0 I think Elon is less predictable and less likely to collaborate with OpenAI while suing them, but I could also see them joining if it’s framed right.
English
0
0
0
18
Joshua Achiam
Joshua Achiam@jachiam0·
I'm going to make a request for some basics from the Pause folks: please outline a practicable version of a pause. Do you mean no training runs above a certain scale? Do you mean furlough the researchers indefinitely? What are you specifically asking for?
David Krueger@DavidSKrueger

A week from today, we will be at Anthropic, OpenAI, and xAI, demanding that leaders agree to a conditional AI pause. These companies are recklessly endangering all of our lives. Their excuse is that they can't pause unilaterally. So they must commit to pausing if others do.

English
41
3
198
43.6K
Austin Wallace
Austin Wallace@austeane·
@MugaSofer @allTheYud @AndyMasley Certainly early Taylor is bad in a very human young adult way, the same way teenagers can be petty and jealous without it meaning they will end up Bad.
English
0
0
1
26
MugaSofer
MugaSofer@MugaSofer·
@austeane @allTheYud @AndyMasley But early-story Taylor is legitimately kiiind of a bad person. She falls in with a gang of antisocial criminals, commits a bunch of crimes to impress them (initially lying to herself that she's totally gonna turn them in), takes over a city & wages a terrorist campaign, etc.
English
1
0
0
20
Andy Masley
Andy Masley@AndyMasley·
Yeah this point does confuse me a lot. Dune is "a cautionary tale about messianic leaders" and yet in later books it's revealed that Paul's actions were the only way to keep humanity free. The messianic leader was in fact smarter than everyone else and had to use cold utilitarian logic to do what was best. He literally is a messianic leader saving everyone. That doesn't seem like the message people often scold you into taking away!
itamar@ItamarLevyOr

@AndyMasley You need to understand that even though paul atreides can canonically see the future, and is following the only course of action that persists humanity, he is actually a bad man.

English
178
101
2.8K
233.5K
Austin Wallace
Austin Wallace@austeane·
Wait, I've heard the arguments on Paul... Do people argue that Taylor was *bad*? I think she's pretty unambiguously good even if she makes "bad choices" and that is one of the main themes of the book. Just curious if the opposite view is sincerely held by a substantial portion of Worm readers!
English
1
0
2
228
Eliezer Yudkowsky
Eliezer Yudkowsky@allTheYud·
@AndyMasley It's bullshit. Of course Paul Atreides and Taylor Hebert are good people! They're in bad situations, but they sacrifice more for other people than 99.9999% of the sniffers ever will. I'm not a telepath for why the authors don't admit this, but I've seen some of the incentives.
English
6
1
59
2.4K
Austin Wallace
Austin Wallace@austeane·
The biggest issue is honestly the creation of the Council, but I genuinely believe even if we don't do a great job, it will still be better in most cases than not having one (here, I think people will fairly disagree) To go point by point on your questions @jachiam0 "What are the conditions under which there should be any collective action towards slowdown? What are the scientific metrics or benchmarks that would decide when to enter or exit a pause?" I believe I answered those well "Are you asking to stop the integration of existing AI into additional systems?" "Are you asking to stop frontier training runs?" "Are you asking for no further tuning of existing models?" "What about incremental progress and minor performance updates, do you want that to stop?" I don't claim to speak for any group, including Pause people, but this is a detailed description of my position, and I would welcome substantial discussion/critique from either direction!
English
0
0
0
13
Austin Wallace
Austin Wallace@austeane·
5/overall This is not near a full pause, but I think it's pareto optimal in a world in which we don't get global cooperation of major AI labs, but do get quite a bit of US collaboration. This proposal disincentizes race conditions between Collaborative labs, without kneecaping those companies, and ideally even giving them a competitive advantage against all non-Collaborative companies. The desired end is that these Collaborative labs together figure out a way to create the most aligned superintelligence we can before a non-Collaborative lab is a major risk of releasing one. I don't think we can stop Superintelligence from being created, and I don't think we can guarantee Superintelligence will be safe, and thus there is a pareto frontier, and the world I describe I believe is: - much closer to that frontier - in the best interest of both the labs and the world - a very difficult, but solvable coordination problem
English
1
0
0
14
Austin Wallace
Austin Wallace@austeane·
@tszzl @_dmca They may be meta useful tokens if Do exploratory work on workflows by seeing what can be done when token usage is deliberately not a consideration, and then figure out efficiency afterwards.
English
0
0
2
633
roon
roon@tszzl·
@_dmca honestly I’m becoming skeptical these are useful tokens
English
31
2
475
23.8K
Austin Wallace
Austin Wallace@austeane·
@Ziiptie Mostly agree though I wish they split him into two! I would love to see an agent focussed on map control and mind games, though that agent shouldn’t also be an entry duelist.
Austin Wallace@austeane

As a day one Yoru main, I think it hits him fairly hard but only for certain play styles, and I'm gonna give a lot of in depth thoughts. It's fine balance wise, but it removes some of "big brain" outplay potential that for me is Yoru's niche. One flash is not a killer, and I think it's fine for Riot to push for more flash initiators. His entry kit is untouched, on purpose I think: He can clone trips, and flash-tp combo to take space. Where he is majorly nerfed for me is fakes on offense, and rotates on defense. I'll give one example play on Haven that I use ~once per ranked: Near start of round, send TP from spawn side of B-Window, one lurker follows TP, while the other four push C/Mid. If we can easily take C, the A becomes a lurk. If they contest C, and the A lurk can secure my TP either because they left site or they win a 1v1, I can tp and immediately plant, and my teammates can harry them from behind. With this change, they are signaling that they want Yoru to be more of an entry duelist. My problem is I think he's worse at that than Neon/Waylay/etc, and I would like to play an agent that directly rewards smart strategic decisions and mindgames and conditioning, and no other agent fits that niche right now. I would have been happiest if they split Yoru into 2: a short-duration tp entry agent with better sentinel disruption; and a long range tp agent without flashes that leans into the mindgames and map control. (Signed a 31 year old with a 7hs% last act D3 solo-queue only)

English
1
0
15
5.1K
ZipTie
ZipTie@Ziiptie·
The TP change is healthy for the agent longterm, being able to rotate to any site twice in a round is just op. It was always the problem and it was the sole reason Yoru is picked in pro play so much. Revert some flash/ult nerfs and Yoru is fine and viable 🙏
VALORANT@VALORANT

x.com/i/article/2033…

English
24
33
1.2K
89.5K
Austin Wallace
Austin Wallace@austeane·
As a day one Yoru main, I think it hits him fairly hard but only for certain play styles, and I'm gonna give a lot of in depth thoughts. It's fine balance wise, but it removes some of "big brain" outplay potential that for me is Yoru's niche. One flash is not a killer, and I think it's fine for Riot to push for more flash initiators. His entry kit is untouched, on purpose I think: He can clone trips, and flash-tp combo to take space. Where he is majorly nerfed for me is fakes on offense, and rotates on defense. I'll give one example play on Haven that I use ~once per ranked: Near start of round, send TP from spawn side of B-Window, one lurker follows TP, while the other four push C/Mid. If we can easily take C, the A becomes a lurk. If they contest C, and the A lurk can secure my TP either because they left site or they win a 1v1, I can tp and immediately plant, and my teammates can harry them from behind. With this change, they are signaling that they want Yoru to be more of an entry duelist. My problem is I think he's worse at that than Neon/Waylay/etc, and I would like to play an agent that directly rewards smart strategic decisions and mindgames and conditioning, and no other agent fits that niche right now. I would have been happiest if they split Yoru into 2: a short-duration tp entry agent with better sentinel disruption; and a long range tp agent without flashes that leans into the mindgames and map control. (Signed a 31 year old with a 7hs% last act D3 solo-queue only)
English
0
0
8
5.6K
Vansilli
Vansilli@Vansilli·
Hmm are Yoru nerfs that RIP? Instead of crossmap rotates for 30 seconds, i can just do mid map rotates for 15. Id be thinking of when id set my TP just like when i choose to activate Dash or Refract on Jett/Waylay. Potentially Yoru's can be even less predictable?
English
9
0
199
22.9K

Keşfet

@theo @AndreaCris74423 @LotharHS @golDonbear @Sathos__voice @jachiam0 @MugaSofer @allTheYud