lauren

4.5K posts

lauren banner
lauren

lauren

@poteto

building @cursor_ai and https://t.co/WDB4U1rwmu. @reactjs compiler core team. prev @meta @netflix

OC Katılım Eylül 2014
1.8K Takip Edilen29.2K Takipçiler
Florin
Florin@florin_dev·
@poteto I'd love to know your pstack setup :D
English
1
0
0
66
Lux Veritas
Lux Veritas@LuxVeritasAeter·
@poteto I would love if it had multiple successful examples.
English
1
0
3
184
lauren
lauren@poteto·
alright im gonna write a new post on tackling big migrations and refactors by yourself
English
24
1
270
14K
lauren
lauren@poteto·
@MountainsGuy1 i always do my writing by hand. writing is thinking and i don't want to outsource that!
English
0
0
8
726
lauren
lauren@poteto·
@mweinbach let me know what you think! easiest way to get started: - run /setup-pstack and choose which models you want - use /poteto-mode whenever you’re doing anything that requires rigor that’s it! the other skills are more situational, the mode skill uses them for you as needed
English
0
0
2
256
Max Weinbach
Max Weinbach@mweinbach·
Cursor with GPT 5.6 Sol is an absolute beast Would love to have it delegate Grok 4.5 subagents to do work honestly, that could be killer
English
34
4
348
21.8K
lauren retweetledi
Cursor
Cursor@cursor_ai·
Introducing side chats, a new way to ask questions and explore ideas without interrupting your main conversation. Each side chat is a durable agent conversation you can @-mention to bring context back into the main thread.
English
205
222
4.4K
393.6K
lauren
lauren@poteto·
so many exciting new models launching this week! so nice that you can use them all in @cursor_ai if you're feeling rich you can get an /arena (from pstack) across fable, sol, and grok 4.5 to work on your hardest hardest tasks
English
13
4
223
8.6K
lauren retweetledi
Elon Musk
Elon Musk@elonmusk·
Good review of Grok 4.5
English
1.1K
2K
10.4K
2.1M
fredrika
fredrika@fredrikalindh·
finally got cursor's first elongated model
fredrika tweet media
English
12
8
554
31.2K
lauren
lauren@poteto·
@stanzillaz if you run /setup-pstack with your own models that won't get changed, so you can run it again later when Grok 4.5 is available in EU!
English
1
0
1
87
Stanzilla
Stanzilla@stanzillaz·
@poteto Grok is not available to me yet, guess I should wait for before I fumble up all the settings with reroutes, yeah?
English
1
0
0
146
lauren
lauren@poteto·
@jessethanley we have something coming soon that should make this easier!
English
2
0
3
139
lauren retweetledi
˗ˏˋ Jesse Hanley ˎˊ˗
Okay so here is my take after merging in ~12 PRs from Grok 4.5 via @cursor_ai into @Bento this morning: It is good and safe to setup as default but only if you have guardrails/not default harness setup. This is the first model from Grok/Cursor you can use for day to day coding, as you were with Opus or GPT 5.5, but you have to have your harness spec'd out to keep it focused and on task. For Cursor users, this means installing @poteto pstack and using the skills to kick off agents ("/poteto-mode build X feature"). If you do that it crushes just as the other models do. If you don't it loses the plot. Example from a PR this morning: customer reported an issue with a timezone picker (kyiv/kiev). With /poteto-mode (pstack): 1 file change, a few lines. Worked. Without skills: 15 file changes, dedicated Bento::Kyiv classes (????), just a mess. Worked but absolute slop. I also find that it has a habit of refactoring aggressively which is sometimes nice but can be annoying as often I just want very targeted fixes. Again, spending time on skills and playbooks seems to be the key to making this model work and stopping it from doing this. I also imagine the ponytail project would also help here greatly. It also says "code judo" a lot. TAKEAWAY: Spend 1 hour today playing with pstack or learning what it is and making your own version then use Grok 4.5 day to day!
˗ˏˋ Jesse Hanley ˎˊ˗@jessethanley

The tasteful latency. The subtle reasoning improvements. Oh my god, look at those Terminal-Bench 2.1 scores.

English
12
1
76
13K
lauren retweetledi
Ramp Labs
Ramp Labs@RampLabs·
Grok 4.5 launched today. It displays frontier-level shell discipline, chaining Bash commands and filtering their outputs more frequently than any other model tested.
English
6
16
278
26K
Mikhail
Mikhail@mikhailmits·
@poteto better than composer in your opinion?
English
1
0
0
34
lauren
lauren@poteto·
it's fast, cheap, and good. first of many more!
Artificial Analysis@ArtificialAnlys

SpaceXAI just released Grok 4.5, and it ranks #4 on GDPval-AA v2 with an Elo of 1543 - behind only the latest Claude releases from Anthropic on real-world agentic knowledge work tasks Grok 4.5 achieved this score at a cost of $0.49 per GDPval task to sit clearly on the Pareto frontier for performance versus cost. This cost is lower than GLM-5.2 and Kimi K2.6, and nearly 90% cheaper than the models ahead of it on our leaderboard. We’re finalizing the remaining Artificial Analysis Intelligence Index evaluations and will share final results soon. Thanks to @SpaceXAI and @elonmusk for their collaboration testing this model ahead of release, and congratulations on the launch!

English
4
3
75
4.9K