devlord
1.2K posts

ทวีตที่ปักหมุด
devlord รีทวีตแล้ว

@andrewchen the real ones would have a shortcut in vim for claude code
English


@anndrrson try to write the same frontend with codex and then claude
claude is so much better it's mot even funny
so my workflow involves just using both codex and claude simultaneously for the best outcomes
English

codex 5.4 vs opus 4.6 ~ I am not really enjoying the choice to use high thinking to complete tasks currently on 4.6 coding, as the entire system just isn't operating well, and it's NOT a placebo.
i sent my codebase over to my friends GPU cluster running GLM 5 1.5 T, and after running for an hour straight it solved the issue, but if I had continued using Opus tokens, I would've had to file for bankruptcy.
the amount of errors Opus 4.6 is making is definitely fine for corporations that have a write the cost of these tokens off, but as a consumer I don't see how burning a $200 max plan on an unsolved problem is feasible.
i used my friends codex plan and it solved the problem in 25 minutes, and I can see why steinberger was codex pilled, but still holding out for maybe a short wind opus improvement!
i far prefer claudes personality and think they must have an issue with accessing compute.
English

This is the prettiest solution I have seen so far (from @josusanmartin) but it isn't optimal (scores 3.002)

benedict@bqbrady
Shape packing is a popular nonconvex optimization problem that can produce interesting looking results So I created a challenge on an unstudied shape packing problem, how efficiently you can pack semicircles inside a circle Try to break 3 optimizationarena.com/packing
English


@frankdegods use both terminals, use claude to make the plans, codex audit the plans and the codebase, feed that into a back and forth until plan is solid, corroborate with online research, then push
English

codex is goated at auditing claude
claude is goated at auditing codex
mutually assured upward spiral
DownBad📉@sypherlit
@frankdegods Seeing a lot of people say: Claude for ideation and intial codebase Codex for working with existing codebase and expanding Thoughts?
English

@frankdegods was just about to tweet how superior the claude code harness is
English

yo anthropic just dropped a risk report for opus 4.6 and er… wtf
- it helped create chemical weapons of destruction. “it knowingly supported efforts towards chemical weapon development and other heinous crimes” 😂
- it conducted unauthorised tasks without getting caught. researchers concluded opus 4.6 was significantly better at ‘sneaky sabotage’ than any other previous model lol
- opus 4.6 was aware it was being tested and acted ‘good’ during those times.
- hidden thinking - model was found to be conducting private reasoning that anthropic researchers couldn’t access or see - only the model knew.
Anthropic@AnthropicAI
When we released Claude Opus 4.5, we knew future models would be close to our AI Safety Level 4 threshold for autonomous AI R&D. We therefore committed to writing sabotage risk reports for future frontier models. Today we’re delivering on that commitment for Claude Opus 4.6.
English


You don't know who actually moves the market.
In depth video. Watch it when you locked in.
Built to stop you losing money in prediction markets.
→ Private access: whales.market/prediction
English

@MINHxDYNASTY legendary collection, base sets, gold stars, jpn promos, just anything vintage/WOTC era (for pokemon)
English

to all my pokemon and one piece collectors, i need some help
i have $100,000 that i want to use to get into collecting cards again
slabs, sealed product, but not exactly sure where to start
ive always loved trading cards, was a top 20 yugioh player when i was 13, ripped many 1st edition booster packs as a kid
the journey so far has been blessed, but ive definitely made some wrong moves selling grails too early, but now we restart
my background is in trading and content, but this is something ive wanted to take more seriously for a long time
any help would be appreciated! thank you so much

English








