Spike Felion

368 posts

Spike Felion banner
Spike Felion

Spike Felion

@FelionSpike

Hey I`m Spike Felion, I have fun making cute and funny vids on youtube. https://t.co/OfAWGk3IoU awooo! stay furry

Katılım Haziran 2020
20 Takip Edilen21 Takipçiler
Willow
Willow@WillowVoiceAI·
Introducing Atlas 1. Willow's new frontier speech-to-text model. It outperforms ElevenLabs, Deepgram, OpenAI, and more by a wide margin. Built on the first scalable, human-powered transcription infrastructure ever built for real-time dictation.
English
246
153
2.3K
809.1K
Spike Felion
Spike Felion@FelionSpike·
@tadikeshwar @SqrRootNegOne @imyouhu So just how it gained that context to begin with when Claude is new with 0 context he can re-read files and boom. Never need to compact. It burns a little tokens reading files every time but TBH I feel it makes Claude work better. Its always fresh Only need a minimal Claude file
English
1
0
2
47
CatFly
CatFly@imyouhu·
Claude Code 源码泄露事件后续越来越精彩了。 有人拿泄露的源码丢给 OpenAI 的 Codex 分析,竟然找到了 Claude Code 疯狂消耗 token 的元凶——autoCompact(自动上下文压缩)机制在失败后会无限重试,完全没有上限。据源码注释记录,曾有会话连续失败高达 3272 次。 修复方法简单到离谱:加一个 MAX_CONSECUTIVE_AUTOCOMPACT_FAILURES = 3 的限制,连续失败 3 次就停止重试。三行代码,搞定。 打完补丁后,这位老哥表示使用额度恢复正常了——之前被吐槽的"用两下就触发限速",很可能有一部分就是这个 bug 在背后偷偷烧 token。 仓库地址放下面了。
Lydia Hallie ✨@lydiahallie

We're aware people are hitting usage limits in Claude Code way faster than expected. Actively investigating, will share more when we have an update!

中文
129
727
6.3K
1M
Spike Felion
Spike Felion@FelionSpike·
@TimTeaFan @krzyzanowskim my understanding is he used chat gpt to rebuild the claude harness but better. he had to use chat gpt because claude models are somewhat "hardcoded / Hardtrained" on not snooping/reverse engineering claude products
English
1
0
2
661
Tim Tiefenbach
Tim Tiefenbach@TimTeaFan·
@krzyzanowskim So this is basically Claude’s agent harness + CLI but in the back runs GPT 5.4?
English
2
0
0
10.5K
Marcin Krzyzanowski
Marcin Krzyzanowski@krzyzanowskim·
I reimplemented "claude" CLI with codex and gpt-5.4-high. It cost $1100 in tokens, and is 73% faster and 80% lower resident memory during sustained interactive use. It is very easy to reverse claude from npm distribution, then reimplement is 1:1. It is indistinguishable from the Anthropic version to the every header and analytics it send back github.com/krzyzanowskim/…
English
164
104
2.1K
1M
Dirtox
Dirtox@reee_ke·
Woof bark guys...😵‍💫 Also, would you be interested to get a render like that? Let me know in the comments
Dirtox tweet media
English
9
31
778
9.7K
Spline
Spline@splinetool·
Today, we're releasing Omma. Create 3D, Websites, and Apps with AI agents. Start now on omma.build Follow @omma_ai
English
180
436
4.2K
988K
🍃Weed Rat🐁➡️ FurTheMore 27’
@Dusky_Boi_ She read it btw ^^ I’ll Keep you updated. I care for her but also really want people to get what they are owed. I myself owe artwork due to medical emergencies this year on my part but I’m working on that rn
English
2
0
8
1.2K
Spike Felion
Spike Felion@FelionSpike·
Idk how people have these struggles, I've casually coded a phone assistant agent with a character and my own programming IDE without touching the code of either. I only have a fuzzy overview of their architectures. Building shit with AI is cool and fun af
Spike Felion tweet media
Luke Parker@LukeParkerDev

im am this close to crashing out. every AI just does dumb stuff unless you are so specific you may as well code. it can help for mass migrations once you've already done the shape and exact impl, and have a bunch of boring work. im so sick of trying to wrangle it lol

English
0
0
0
41
Team Reagent
Team Reagent@Reagent_Systems·
its happening
Team Reagent tweet media
English
1
0
1
43
Spike Felion retweetledi
Torvid
Torvid@Torvid21·
I made a tool that bakes NormalizedCrow's FeathersAndFur Shader into the main mesh. torvid.net/featherbake This makes it work in fallback, with other shaders like poiyomi, improves performance, and lets you export to external software like Blender!
English
10
69
871
24.5K
The Fox Nest
The Fox Nest@TheFoxNestStore·
Furry made Femboy outfit are back for a round of preorder 👀 Don’t miss it :3 🦊
The Fox Nest tweet media
English
2
18
625
6.8K
Reethu
Reethu@ritu_twts·
as a dev, what was your first code editor?
Reethu tweet media
English
874
52
1.3K
103.9K
jolly ass enjoyer 18+
jolly ass enjoyer 18+@ArizonaTheVixen·
@FelionSpike @SnooferX3 @dashydoggo Most people are actually working hard, working 40h a week minimum wage to have nothing left at the end of the month but crippling debt, i can't even imagine being in this position, and still having the strenght to continue And it's most people reality
English
1
0
1
81
dashy 🔜 FWA, AC?, MP
dashy 🔜 FWA, AC?, MP@dashydoggo·
i lose a year off my life every time i see people in the fandom starting adulthood in FL or Cali with the lowest possible difficulty setting on like how are you a 18 y/o with a muscle car and a shoreline zip code while attending a private uni
English
58
33
851
32.3K
Spike Felion
Spike Felion@FelionSpike·
@RedPixelOne @rantingjax Still sucks ass but hopefully that clears up that confusion. Gotta just make the things u want to make for the love of making it. Because when it comes to audiences sometimes they will like the simple content more then quality. Have had friends with the same scenario on YouTube.
English
0
0
0
8
Spike Felion
Spike Felion@FelionSpike·
@RedPixelOne @rantingjax It's the same thing though. If less people are interested in a post. Which less people are because it's not as thirsty and furries are too thirsty. Then it will show that post to less people as a result. It's only showing the post to less people because less people like it
English
1
0
0
15
RedPixel One
RedPixel One@RedPixelOne·
This is why i start to really tire of social media. the post on the left took 10 mins to render 5 mins to set up 15. mins. got over 5k views, post2 took 1 hr to set up, 2 hrs to render the finalized version 3 hrs Total a bit over 370 views. i like the 2nd one 100 times more
RedPixel One tweet mediaRedPixel One tweet media
English
35
12
195
13.5K
Spike Felion retweetledi
JT
JT@JTwusky·
People do realize they could also just make small fun trips with their friends instead of cons right? No worries about hotel availability, spend quality time instead of quickly bumping into friends then never see them the rest of the weekend? Small trips are just as fun imo
English
36
47
793
30.3K
alex fazio
alex fazio@alxfazio·
pewdiepie just trained his own llm, and it beats gpt-4o on coding benchmarks. an apocalyptic, civilization-ending catastrophe of laughably, cosmically disproportionate magnitude for the entire ml research job category
alex fazio tweet media
English
143
361
7.7K
1.1M
Team Reagent
Team Reagent@Reagent_Systems·
mom, am i a data scientist now?
Team Reagent tweet media
English
1
0
3
29