Andrew Margetts @CodeVicious

🚀 Entrepreneur on a mission to innovate | Ethical Hacker safeguarding the digital world | 💡 Always learning. South Africa Joined February 2018

Tweets

591
Followers

61
Following

943
Likes

1K

Lisan al Gaib @scaling01

3 days ago

Opus 4.8 just broke ARC-AGI-3 it tripled GPT-5.5's score we are now at a breathtaking 1.5% human efficiency

99 74 2K 174K 265

View Details

Theo - t3.gg @theo

a week ago

Cursor has updated CursorBench with Opus 4.8. It is more efficient, but performs slightly worse than Opus 4.7 within margin of error.

108 44 1K 149K 132

View Details

Max For AI @MaxForAI

a week ago

笑死了，Claude Opus4.8蒸馏了阿里巴巴Qwen啊🤣 通过API用中文问你是谁，会很大概率回答我是通义千问（Qwen），是阿里巴巴集团旗下的统义实验室自主研发的超大规模语言模型。

502 134 2K 927K 441

View Details

Uber blew through its entire 2026 Claude Code budget by April and is slowing hiring to cover the API bill. The COO admits the massive token burn isn't translating to shipped features. We didn't automate engineering. We just traded actual headcount for a while-loop with a corporate card.

111 408 6K 370K 720

View Details

Tuomas Artman @artman

4 weeks ago

Today is a hard day. I shared this note with the @linear team today: We’ve made the difficult decision to increase our workforce. This is not a cost-cutting exercise or a reflection of anyone’s performance. We’re simply reimagining every role for the agentic AI era. We’re hiring. We’re sorry about that.

449 640 14K 979K 1K

View Details

Austin Barnhill @bh30317

4 weeks ago

Pretty sure this is what ⁦@brian_armstrong⁩ means when he says, “Non-technical teams are now shipping production code and many of our workflows are being automated.” Keep up that good workflow ⁦@coinbase⁩ 🫡

121 193 5K 326K 226

View Details

Andrew Margetts @CodeVicious

4 weeks ago

@Nicolascole77 Social

0 0 0 5 0

View Details

Jason Fried @jasonfried

a month ago

The faster people can get to something 80% done and unstable, the more they are going to crave stuff that's already 100% done and reliable.

65 107 1K 81K 272

View Details

beginbot 🃏 @beginbot

a month ago

I told my gf I can't hang out right now Github is up, so I have to work don't know when I'll get this chance again

44 391 9K 185K 197

View Details

Benjamin Todd @ben_j_todd

a month ago

Hard to think of a more AI-exposed job than Filipino call centres, but employment was up 4% in 2025.

26 47 753 96K 90

View Details

Alex Imas @alexolegimas

a month ago

@copperchunk The way this works is if you want me to take the time to read it you need to take the time to write it.

14 22 886 77K 70

View Details

staysaasy @staysaasy

a month ago

The uncertainty that AI brings is causing management teams to lose their heads strategically, tactically, and operationally. I have never seen anything like it. There's some humiliating behavior going on out there.

27 49 660 36K 79

View Details

Nikita Bier @nikitabier

a month ago

One of the most important things that Elon taught me: People engaging in fraud are always the first and loudest to complain.

6K 4K 47K 3.2M 2K

View Details

Karri Saarinen @karrisaarinen

a month ago

A common dynamic I observe with AI: it feels most impressive when you don’t know much about the subject, don’t care or don’t have a clear idea of what the you want. This applies across design, code, legal, and more. If I don’t know code very well, every piece of code it writes feels very impressive. Once you know what something should feel or look like, it becomes almost impossible to guide AI there. And you definitely can’t one-shot it.

258 392 3K 571K 849

View Details

Yuval Adam @yuvadm

a month ago

1/ Project Eleven just awarded 1 BTC for "the largest quantum attack on ECC to date", a 17-bit elliptic curve key recovered on IBM Quantum hardware. I replaced the quantum computer with /dev/urandom. It still recovers the key.

28 143 841 189K 229

View Details

Cairo Smith @cairoasmith

a month ago

China ten years ago: wow I can't believe all these equatorial countries are agreeing to payday advance loans with 63.8% APR, I'm going to be so rich when I collect.

Wall Street Journal Opinion @WSJopinion

a month ago

Beijing wants to use the International Monetary Fund to rescue the array of distressed Chinese loans around the world. on.wsj.com/3Qqe5JZ

180 341 2K 2.1M 566

75 812 16K 913K 1K

View Details

gabe @allgarbled

a month ago

If you asked Claude to build you a house it would build you the most beautiful house, and then you’d go inside and you’d be like, “Claude there’s no bathrooms.” And Claude would say, “There were no bathrooms before either, so it’s actually a pre-existing issue”

40 139 5K 201K 219

View Details

Dan ⚡️ @d4m1n

2 months ago

so let me get this straight > OpenAI renamed their whole division "AGI Deployment" > Jensen said AGI is "already in the room" and then ARC-AGI-3 drops and on a scale from 0 to 100% SOTA models score: GPT-5.4: 0.26% Gemini Pro: 0.37% Claude Opus 4.6: 0.25% Grok: literally 0% meanwhile 100% of human testers solved every single environment. first try. no instructions. no training. AGI is "in the room" brother it couldn't find the room

143 176 3K 251K 583

View Details

Lisan al Gaib @scaling01

2 months ago

not gonna lie this is and looks terrible for anthropic

wh @nrehiew_

2 months ago

The 4.7 tokenizer treats whitespace as separate tokens? A string consisting of 50 one-token words separated by Whitespace tokenizes to ~50 more tokens than with the 4.6 tokenizer. If so, the 1.35x more token estimate seems way too low.