Thomas Lancer

Building

[Thomas Lancer @LancerThomas]

@maverickecom agent

[Thomas Lancer @LancerThomas]

@AlfieJCarter cold

[Thomas Lancer @LancerThomas]

@FoundersPodcast Have you read Revolution in the valley? Andy hertzfeld wrote it. first person account of the making of the Mac.

Also he has a site called folklore (dot) org with a bunch of stories from the Mac team themselves. V cool

[Thomas Lancer @LancerThomas]

Churchill during ww2 on what he wanted for Britain.

Incredible that the freedoms he fought to defend in the face of a takeover by a Nazi regime are now being taken away by Britain’s own leadership.

[Thomas Lancer @LancerThomas]

@ericzakariasson @cursor_ai specific files - anytime its touching DB model files include the rule on SOP for running alembic migration

also for all .tsx files have a rule that tells it to separate business logic from UI code since it loves combining everything into one file if you don't prompt it not to

[Thomas Lancer @LancerThomas]

@layer07_yuxi what paper/book etc is the original quote from? looks interesting

[Thomas Lancer @LancerThomas]

@ryolu_ Would be great to have the ability to manually tell the agent that it can move onto the next step when running terminal commands

if you try to open an SSM session or SSH into an instance it'll execute the command to do so, but then it gets stuck here @ryolu_

[Thomas Lancer @LancerThomas]

bet you could run a very successful vc fund with one simple rule: never pass up investing in a founder who is under 25 and balding

[Thomas Lancer @LancerThomas]

does anyone know someone with experience in high volume youtube scraping ? (downloading 1,000s - 10,000s+ vids/hr)

working on cool experiment that requires this - willing to pay for an hour of their time - pls dm if you know someone!

[Thomas Lancer @LancerThomas]

@scaling01 This won’t age well

[Thomas Lancer @LancerThomas]

@davidfigeira 10k

[Thomas Lancer @LancerThomas]

Goes hard

[Thomas Lancer @LancerThomas]

one interesting way to improve performance on coding w/LLMs is to have it generate 3-5 responses at same time then have another model review it's answers and select best one (or ideally be able to test all it's solutions in parallel somehow and choose the one that works)

i find myself doing this already when solving bugs

i'll open 3 grok windows w/think mode, 2 o3-mini-high windows, 2 o1-pro windows and paste the file and context in all of them

oftentimes 2-3 out of the 7 will solve the issue and the other ones won't

[Thomas Lancer @LancerThomas]

code always does exactly what you tell it to do. that is also the problem sometimes when what you told it to do was dumb 🤣

[Thomas Lancer @LancerThomas]

@shaoruu @cursor_ai Being able to direct the apply model based on line number. In files over 1,500 lines it really struggles to find right spot to apply but if I could give the specific line numbers to start at or it could figure that out on it’s own would make the apply model way more accurate

[Thomas Lancer @LancerThomas]

Feels great to be an American

[Thomas Lancer @LancerThomas]

Incredibly impressive