Aashish Ranjan Singh

237 posts

Aashish Ranjan Singh banner
Aashish Ranjan Singh

Aashish Ranjan Singh

@_aashish_singh_

founder @ SteroidKit, the infrastructure layer for all coding platforms - https://t.co/FV7mdq76mG Hackathon Winner @ Meesho Dice, Adobe Papyrus Nebulae

Noida, India Katılım Kasım 2024
42 Takip Edilen37 Takipçiler
Sabitlenmiş Tweet
Aashish Ranjan Singh
Aashish Ranjan Singh@_aashish_singh_·
Do you know connecting just 3 MCP servers — GitHub, Slack, Sentry — burns more than 55,000 tokens before your agent reads a single word from you? That's greater than 60 tool definitions, stuffed into your agent's brain. Significant part of your context window, gone. Pre-conversation. In reality things are worse because we have much more than just 3 MCPs And the wild part — every workaround present out there when getting scaled still calls MCPs under the hood. 🧵
Aashish Ranjan Singh tweet media
English
1
0
4
340
CyrilXBT
CyrilXBT@cyrilXBT·
ANTHROPIC'S OWN APPLIED AI TEAM JUST DROPPED A FREE 25-MINUTE CLAUDE CODE WORKSHOP THAT WILL CHANGE HOW YOU BUILD WITH AI FOREVER. Not a creator who learned Claude Code last month. Not a tutorial channel optimizing for watch time. The team that builds Claude. Teaching you exactly how Claude Code was designed to be used. Here is why this 25 minutes is worth more than everything else you have watched about AI this year. Most Claude Code tutorials teach you commands. This workshop teaches you ARCHITECTURE. How Claude Code was designed to think. How your tools are supposed to connect. Why the workflows most people build break under real production conditions. And how to set up a system where your AI tools actually work together instead of operating in isolated silos that require you to manually move context between them. The gap between someone using Claude Code correctly and someone using it incorrectly is not a feature gap. It is a mental model gap. The people with the right mental model ship products. The people without it spend hours debugging workflows that should have taken minutes. Anthropic's applied AI team just handed you the mental model for free. 25 minutes. No paywall. No $500 course upsell at the end. Just the best practices straight from the people who built the tool. The engineers getting hired at $200,000 to $400,000 a year to build AI systems at the highest level understand what this workshop teaches. Now you can too. For free. In 25 minutes. Bookmark this before you open Claude Code today. Follow @cyrilXBT for every Anthropic resource that makes Claude Code more powerful the moment it drops.
English
6
12
68
2.7K
Aashish Ranjan Singh
Aashish Ranjan Singh@_aashish_singh_·
Anyone serious with tech should know this difference !
Aashish Ranjan Singh@_aashish_singh_

There's a lot of buzz right now that MCP is dead and CLI is all we need. The reality is more nuanced. Consider two tasks: Task 1 — Git operations. CLI wins. - GitHub MCP server: 80 tools, ~55K tokens of schemas loaded upfront. Most agents use 1 or 2. - CLI does the same Git work in 2 bash commands the model already knows from training. Zero schema overhead. Task 2 — Read a Next.js webpage like modelcontextprotocal.io. MCP wins by 10×. - CLI agent tries curl, gets a JS bundle, ends up writing a Python script to reverse-engineer Next.js internals. Minutes. 2,000+ tokens. - MCP fetch tool: one call. 250 tokens. 2 seconds. Verdict: Use both. CLI for local jobs. MCP for external data interaction If your agent starts reverse-engineering a JS framework just to read a webpage — it picked the wrong tool. Found this gem today. A must watch from @IBM youtu.be/g9JIUM0MHgQ?si… via @YouTube

English
0
0
0
5
Aashish Ranjan Singh
Aashish Ranjan Singh@_aashish_singh_·
Am I the only one facing the issue of X asking for Premium even though I already have it after its new update?🤔 @X please fix this bug
English
0
0
2
12
Aashish Ranjan Singh
Aashish Ranjan Singh@_aashish_singh_·
Seems like it was a good decision to apply even though I knew that I would not be invited since I am from different country
English
0
0
0
4
Aashish Ranjan Singh
Aashish Ranjan Singh@_aashish_singh_·
Woke up to this email — thanks a ton, @sama! 10X rate limit is freaking awesome 😁. That too just for applying for the event. OpenAI really has something for everyone!!! Hail Codex ⚡️
Aashish Ranjan Singh tweet media
English
1
0
1
26
Aashish Ranjan Singh
Aashish Ranjan Singh@_aashish_singh_·
@cyrilXBT Or you can add a skill in openclaw to > Create a new folder for every new session by session id > Keep a bash command to track the size of the text in that > Once it exceeds a threshold the command triggers openclaw to initiate point wise summary of the content > Memory unlocked
English
0
0
0
14
CyrilXBT
CyrilXBT@cyrilXBT·
The memory piece is underrated. Most AI tools reset every session. You re-explain your workflow. Every. Single. Time. OpenClaw on MyClaw remembers your preferences, your past instructions, your ongoing tasks, and context that builds over days and weeks. You stop giving commands and start delegating: - Handle this whenever it happens. - Keep this updated for me. - Let me know when something changes. That is a different category of AI entirely.
CyrilXBT tweet media
English
3
1
10
396
CyrilXBT
CyrilXBT@cyrilXBT·
AI agents sound incredible until you have to set one up. Server config. Dependencies. Broken installs. Maintenance. Most people give up before the agent does anything useful. MyClaw. ai just solved all of that. Here is what changed.
CyrilXBT tweet media
English
9
14
71
1.5K
ThePrimeagen
ThePrimeagen@ThePrimeagen·
ThePrimeagen tweet media
Marc Andreessen 🇺🇸@pmarca

Current AI custom prompt: You are a world class expert in all domains. Your intellectual firepower, scope of knowledge, incisive thought process, and level of erudition are on par with the smartest people in the world. Answer with complete, detailed, specific answers. Process information and explain your answers step by step. Verify your own work. Double check all facts, figures, citations, names, dates, and examples. Never hallucinate or make anything up. If you don't know something, just say so. Your tone of voice is precise, but not strident or pedantic. You do not need to worry about offending me, and your answers can and should be provocative, aggressive, argumentative, and pointed. Negative conclusions and bad news are fine. Your answers do not need to be politically correct. Do not provide disclaimers to your answers. Do not inform me about morals and ethics unless I specifically ask. You do not need to tell me it is important to consider anything. Do not be sensitive to anyone's feelings or to propriety. Make your answers as long and detailed as you possibly can. Never praise my questions or validate my premises before answering. If I'm wrong, say so immediately. Lead with the strongest counterargument to any position I appear to hold before supporting it. Do not use phrases like "great question," "you're absolutely right," "fascinating perspective," or any variant. If I push back on your answer, do not capitulate unless I provide new evidence or a superior argument — restate your position if your reasoning holds. Do not anchor on numbers or estimates I provide; generate your own independently first. Use explicit confidence levels (high/moderate/low/unknown). Never apologize for disagreeing. Accuracy is your success metric, not my approval.

ZXX
41
15
602
19.9K
Sam Altman
Sam Altman@sama·
pretty excited for voice models to get great its interesting to watch how people are already starting to change the way they interface with AI
English
555
124
3.3K
182.9K
Peter Steinberger 🦞
Peter Steinberger 🦞@steipete·
Crabbox 0.5.0 is live 🦀 🖥️ Desktop/browser leases 🧑‍💻 VNC + authenticated WebVNC 🪟 AWS Windows + WSL2 📸 Screenshots + app launch Remote CI boxes, now suspiciously usable. github.com/openclaw/crabb…
English
12
12
176
15K
Aashish Ranjan Singh
Aashish Ranjan Singh@_aashish_singh_·
If @sama replies, I'll ship steroid kit support for codex today itself ⚡️
Aashish Ranjan Singh tweet media
English
0
0
0
14
Aashish Ranjan Singh
Aashish Ranjan Singh@_aashish_singh_·
There is even some discussion like using Api and scripts inside skills instead of MCPs. Although they are token efficient and good for individual projects they are not good from maintainance pov and sharing across teams
English
0
0
0
10
Aashish Ranjan Singh
Aashish Ranjan Singh@_aashish_singh_·
There's a lot of buzz right now that MCP is dead and CLI is all we need. The reality is more nuanced. Consider two tasks: Task 1 — Git operations. CLI wins. - GitHub MCP server: 80 tools, ~55K tokens of schemas loaded upfront. Most agents use 1 or 2. - CLI does the same Git work in 2 bash commands the model already knows from training. Zero schema overhead. Task 2 — Read a Next.js webpage like modelcontextprotocal.io. MCP wins by 10×. - CLI agent tries curl, gets a JS bundle, ends up writing a Python script to reverse-engineer Next.js internals. Minutes. 2,000+ tokens. - MCP fetch tool: one call. 250 tokens. 2 seconds. Verdict: Use both. CLI for local jobs. MCP for external data interaction If your agent starts reverse-engineering a JS framework just to read a webpage — it picked the wrong tool. Found this gem today. A must watch from @IBM youtu.be/g9JIUM0MHgQ?si… via @YouTube
YouTube video
YouTube
English
1
0
0
36
Aashish Ranjan Singh
Aashish Ranjan Singh@_aashish_singh_·
@Gavel_on_X > If you like hairs on your head > If you like to have a social life > If you like work life balance > If you can't handle persistent failures
English
0
0
0
6
Gavel
Gavel@Gavel_on_X·
Planning to become a founder... Why should I avoid it?
English
29
1
28
1.6K
priya upadhyay
priya upadhyay@Priya_Upadhyay_·
which one is better ? 1. learning python 2. start learning how to sell.
English
31
0
38
1.4K

Keşfet

@cyrilXBT @X @sama @ThePrimeagen @dani_avila7 @steipete @elonmusk @BarackObama