tadpole

8.5K posts

tadpole banner
tadpole

tadpole

@cryptoanuran

Katılım Ağustos 2017
1.5K Takip Edilen1.4K Takipçiler
Fishy Catfish
Fishy Catfish@CatfishFishy·
I'm dropping a thread of all the protocols that had to freeze their interop because of LayerZero being compromised. Let's go:
English
133
307
1.6K
325.4K
tadpole
tadpole@cryptoanuran·
@trq212 tyft /clear is literally just “new session, same window”? For plans: when a plan gets accepted does context get cleared or compacted at all or is everything still present? Is plan mode designed such that accepted plans are should generally be handed off to a new session?
English
1
0
0
178
Mayne
Mayne@Tradermayne·
More and more people are getting cooked by the 1099-DA. This is above your standard CPAs paygrade and is definitely not DIY territory. This article is worth a read if you want to know what's going on. cointracker.io/blog/the-1099-… I use @countonsheep and they will give you are free consult if you book thru this link. countonsheep.com/tradermayne
English
29
15
182
31.7K
doublecrxss
doublecrxss@doublecrxss·
@TheEbonyMaw I can’t stand the road trip episode for similar reasons. Chili is the biggest obstacle the whole time but the show treats it like Bandit is the unreasonable one.
English
3
0
23
4.3K
Maw
Maw@TheEbonyMaw·
Me watching Bluey initially as a brand new parent: “Bandit & Chili are so great. Model parents. I hope to be like them one day.” Me watching Bluey for years & after being a parent for 6 years: “JUST SAY NO TO YOUR CHILDREN FOR ONCE IN YOUR LIFE YOU WALKING DOORMATS!!”
English
70
266
10.5K
365.5K
tadpole
tadpole@cryptoanuran·
@trq212 @dexhorthy You guys have the skill-creator skill (why a plugin and not packaged with claude code though?), but a skill-improver/fixer skill for existing skills would be fantastic that implements all current best practices
English
0
0
0
63
Thariq
Thariq@trq212·
@dexhorthy yeah their weakness is that its too easy to make bad ones
English
8
0
49
5K
Thariq
Thariq@trq212·
somehow skills are still underrated
English
279
67
1.9K
147.4K
tadpole
tadpole@cryptoanuran·
@TheZvi image understanding and search it still stands out Gotta believe they’re cooking hard, or something is just going completely awry over there
English
0
0
0
69
Zvi Mowshowitz
Zvi Mowshowitz@TheZvi·
Do we still have any Gemini 3 Pro fans out there? Or is this fully a two-horse race right now between 5.3-Codex and Opus 4.6?
English
129
2
321
46.1K
tadpole
tadpole@cryptoanuran·
Bitcoiners civil war’d over something as paltry as blocksize and segwit in 2017, where’s the damn quantum hardfork war?? C’mon miners get active
English
1
0
1
91
Michael Saylor
Michael Saylor@saylor·
₿uilt for the Long Run.
Michael Saylor tweet media
English
3K
1.8K
27.1K
1.6M
qw
qw@QwQiao·
2nd best way to get exposure to anthropic is to buy amzn (primary cloud provider) best way is to get private shares on @PreStocks
English
50
36
370
223.7K
tadpole
tadpole@cryptoanuran·
@JedKingly Should really start pushing norms to keep Christmas lights up all winter. LED lighting is cheap to run, might as well keep a little color amidst the 2 months of gray post-Dec25
English
0
0
5
209
Kingly
Kingly@JedKingly·
Winter is great because 2 months in you’re like that was neat but that’s enough of that. And then Actual Winter hits.
English
206
12.4K
194.9K
2.5M
tadpole
tadpole@cryptoanuran·
@aaxsh18 @rahulgs not perfect but could caption every image file/make sure it has thorough tagging in metadata (yes then the problem becomes how to caption/tag-enrich every png on your computer)
English
1
0
0
52
Aamir
Aamir@aaxsh18·
@rahulgs I think it's going to be interesting how things will move once we are hitting truly long running agent or move beyond code (like Claude cowork). We will need better (multimodal) search etc
English
2
0
5
817
rahul
rahul@rahulgs·
rag isn't dead it's just harder than agentic search there is a case for agentic being better than semantic search but claude code's arch doesn't really let it do semantic search anyway 1. claude code basically consumes the standard anthropic /v1/messages api 2. anthropic doesn't offer any embedding models 3. building an vectordb/index in the cloud would be tricky with data retention/code uploads, especially with enterprise. also more complexity the current client-only claude code arch doesn't really leave anthropic a choice, even if agentic search works better. agentic search is just so much simpler to implement, all you need is grep
Theodor Marcu@theodormarcu

I'm glad to see other serious builders in the industry catch up to the fact that RAG is basically dead Fun fact: this insight formed the basis of our intuition when we trained SWE-grep and built Fast Context with it Traditionally, agent builders had two options for context retrieval: - Embedding Search (RAG): fast, but imprecise - Agentic Search: slow, but precise It felt like an inescapable tradeoff - you could choose either slow + smart or fast + dumb So we trained a model (SWE-grep) and built a third option: fast + smart. That's what powers Fast Context in Windsurf

English
26
12
279
53.1K
tadpole
tadpole@cryptoanuran·
Google’s gotta be cooking on new Gemini 3 preview snapshots right? Been two months and Claude code has completely stolen the discourse Been almost two months and we are still here
tadpole tweet media
English
0
0
0
72
tadpole
tadpole@cryptoanuran·
@thinkingshivers @Zinnin Same thought here
tadpole@cryptoanuran

@thinkingshivers Been loving the game, thanks for the daily dopamine. Could you add a normalized score to the Share text so there’s more granularity than medals alone? Raw score can be a soft spoiler, but a normalized % (best for that day = 100%) would allow ranking without revealing the target

English
0
0
1
30
Shivers
Shivers@thinkingshivers·
@Zinnin Seems like a pretty good idea. Let me think about it.
English
1
0
1
182
Shivers
Shivers@thinkingshivers·
I made a simple puzzle game! Try to enclose the horse in the largest possible area with limited walls.
English
345
1.7K
30.1K
2.3M
tadpole
tadpole@cryptoanuran·
(Speaking as an apparently top 1% chat user if the 2025 year end thing is true)
English
0
0
0
40
tadpole
tadpole@cryptoanuran·
I still stick w ChatGPT for personal use since I never need to worry about rate limits for 5.2 thinking (I just use plus for personal use). Last time I used Claude.ai rate limits were bad. I’d only switch if I could talk to opus 4.5 thinking with no rate woes
English
1
0
0
215
tadpole
tadpole@cryptoanuran·
Pretty cool that in 2026 the big 3 all have good enough models (Gemini 3 pro, gpt 5.2, opus 4.5) and you can safely choose a main chat app on “taste”, or features Agree with this take generally. Chatgpt 5.2 is fairly sterile but useful. Talking to opus more interesting
will brown@willccbb

my biggest holiday LLM revelation was that Opus is just a magnificent chat model, far better than anything else i’ve ever tried. swapped from ChatGPT to Claude as daily chat app. finding myself asking way more & weirder questions than i ever asked Chat, and loving it

English
2
0
1
539
tadpole
tadpole@cryptoanuran·
@thinkingshivers Been loving the game, thanks for the daily dopamine. Could you add a normalized score to the Share text so there’s more granularity than medals alone? Raw score can be a soft spoiler, but a normalized % (best for that day = 100%) would allow ranking without revealing the target
English
2
0
7
373
Shivers
Shivers@thinkingshivers·
Added the ability to report levels. Not the coolest feature, but it was necessary; a bug allowed walls to sometimes be placed over cherries, resulting in impossible levels/scores. Now impossible (or inappropriate) levels can be flagged.
English
18
10
1K
56.6K
doomer
doomer@uncledoomer·
a fun game to play with foid when you take her for some Target™️ time is "find me baby formula that doesnt have seed oils in it" its an unwinnable game. they all, by law, in america, have to have seed oils in them. its one of the most evil things ive ever seen
doomer tweet media
English
86
344
5.1K
211K