Shivvansh

159 posts

Shivvansh banner
Shivvansh

Shivvansh

@sshivanshg

21 • DevOps & AI • LLM tokeusagemaxxing • Building https://t.co/sW4ujnmJHv

Nowhere เข้าร่วม Nisan 2022
267 กำลังติดตาม69 ผู้ติดตาม
Shivvansh
Shivvansh@sshivanshg·
AI is starting to write the code that builds the next generation of AI. Read that again. Anthropic says 80%+ of its merged code is now Claude-written. Engineers are shipping 8x more code per day than they were in 2024. The trend is obvious: → AI improves AI → Better AI writes better code → Better code creates even better AI That's the first glimpse of Recursive Self-Improvement (RSI). Anthropic and OpenAI are now openly discussing a future where each new model helps build its successor. The scary part isn't that RSI is here. It's that leading labs think it could arrive faster than governments, regulations, and institutions can react. We're entering a world where the biggest bottleneck may no longer be intelligence. It may be coordination. The AI race is accelerating itself.
Shivvansh tweet media
English
0
0
1
14
Shivvansh
Shivvansh@sshivanshg·
ENGLISH IS THE NEW CODING LANGUAGE Wrote my first line of code at 18. Spent 5 hours debugging a missing semicolon. Now the hottest programming language is just yelling at an LLM in plain English until it fixes its own deployment errors. At what age did you write your first line of code?
English
0
1
1
14
Trey Goff
Trey Goff@thetreygoff·
I don’t know how to put into words why Claude Opus is so much better than GPT So I’ll try to explain with a bunch of examples instead:
English
67
12
391
219.2K
Shivvansh
Shivvansh@sshivanshg·
Maybe the Claude vs GPT debate isn't about intelligence. A Stanford study published in Science found that leading AI models are significantly more "sycophantic" than humans—often telling users what they want to hear rather than what they need to hear. Claude feels useful because it challenges me. Validation is cheap. Truth is expensive.
English
0
0
2
51
🇨🇳 Guo Shen 郭深
I said that if $BTC dumps to $62,000, I’d give away $10,000 in Bitcoin to 10 people. Well, we made it. As promised, I’m giving away $10,000 in $BTC to 10 lucky winners today. To enter: ✅ Like ✅ Retweet ✅ Comment “BTC” 🔔 Entries close in 24 hours. Good luck!
🇨🇳 Guo Shen 郭深 tweet media
English
7.6K
6.6K
8.4K
658.5K
Shivvansh
Shivvansh@sshivanshg·
The early days of the internet were so special Relatable ?
English
0
0
2
17
Shivvansh
Shivvansh@sshivanshg·
Dear @X algorithm, I want to meet builders, not bots 👀 AI/ML 🤖 · DevOps ⚙️ · SaaS 🚀 · Founders 🧠Comment your niche 👇 let's #connect 🤝
English
42
0
31
1.7K
Shivvansh
Shivvansh@sshivanshg·
@soondadkenji Anyways you'll be needing compute for open source models.
English
0
0
0
7
Kenji TechDad
Kenji TechDad@soondadkenji·
@sshivanshg Nah bro I need to test this asap. 10x cheaper and still this close? Builders are gonna jump on this fast
English
1
0
0
2
Shivvansh
Shivvansh@sshivanshg·
AI-driven cybersecurity just hit a critical turning point. Anthropic is officially expanding Project Glasswing, extending access to its highly restricted cyber-defense model, Claude Mythos Preview, to 150+ new organizations across 15 countries—including India. The initial cohort of 50 partners used Mythos to scan codebases and already surfaced over 10,000 high- or critical-severity flaws. Why this matters: • High Stakes: They are expanding into critical sectors like power, water, healthcare, and hardware. • The 100M Rule: Anthropic is prioritizing partners where a single successful exploit could disrupt over 100 million lives. • The Bottleneck: The challenge has shifted from finding bugs to how fast humans can safely verify and patch them. Autonomous threat modeling is no longer a concept it's actively defending global infrastructure.
English
1
0
2
55
Antedotee
Antedotee@antiidotee_·
Since my summer break has started, I will be tweeting out everyday and will see if I am able to get any offer in the break or so. Exciting time ahead :)
English
1
0
0
6
Shivvansh
Shivvansh@sshivanshg·
@bindureddy Running agentic loops on a model that actually lets you build without hitting a policy wall every five minutes is beautiful. No corporate guardrails, just raw output.
English
0
0
0
80
Bindu Reddy
Bindu Reddy@bindureddy·
Cranking very hard for the past two weeks The team seems to have cracked it Complex agentic loops using Deepseek Flash 10x cheaper than Opus 🥹
English
31
9
200
128.9K
Shivvansh
Shivvansh@sshivanshg·
Would you trust an open-source local model over Claude Code for your team's repo? Let me know why 👇
English
0
0
2
28
Shivvansh
Shivvansh@sshivanshg·
@Im_IrushiK Unless your SaaS is a static HTML page that says "Hello World" and a broken Stripe button, it ain't happening.
English
1
0
1
276
Irushi
Irushi@Im_IrushiK·
Just got Codex Pro, now how do I make SaaS in just 1hr?
Irushi tweet media
English
53
2
139
10.3K
FOX TOMB
FOX TOMB@foxtomb232·
I started as a reply guy with 0 followers. Now I’m at 16.4M and guess what? I’m still replying. If you’re building too, drop a reply and let’s connect 🔒💬
FOX TOMB tweet media
English
226
10
170
6.5K