Sumit Parakh

Today I mark ~2.5 years of my journey into Open Source contribution. I must say that this decision was one of the best I've ever made. I learnt a LOT from this and met some amazing people from OSS community. My gratitude to all OSS folks who are directly and/or indirectly. 1/5

Before you form an opinion, experience it. It is easy to jump to conclusions and form opinions based on what others say. But nothing beats the firsthand experience. Before forming a strong opinion, take the time to see and understand things for yourself. When you actually go through something, you see the full picture - the good, the bad, and everything in between. That's when your opinion truly holds weight. Holds true for all things work and all things learning. If you are trying to make an impact, challenge the status quo, participate in design discussions, mentor people, make decisions, or navigate uncertainty - experience matters, because real impact demands real understanding. The best part? It holds true for life, too.

We hired a junior developer to write the simple code, so we don't have to spend a ton of money on tokens for those basic/primitive tasks

The pricing collapse in AI is the most underreported story of 2026. Six months ago, running a model that competes with GPT-5 and Claude Sonnet required an API subscription and recurring costs. Today, Alibaba released an open-source model that beats both on multiple benchmarks and runs on a consumer laptop. Qwen 3.5 ships with 397 billion total parameters but only activates 17 billion per token. It supports 201 languages, a million-token context window, native multimodal processing, and costs about 60% less to run than its predecessor. On benchmarks, it matches or exceeds GPT-5.2 and Claude Opus 4.5 on coding and reasoning tasks while being a fraction of the size. The API pricing comparison tells the story clearly. Claude Opus 4.6 runs $5 input and $25 output per million tokens. GPT-5.3 Codex runs $1.75 and $14. Qwen 3.5 Plus runs $0.40 and $2.40. That's not a marginal difference. That's a structural shift in who can afford to build with frontier-level AI. And Qwen isn't alone. GLM-5 from Zhipu AI, a 744 billion parameter model, scores 77.8% on SWE-bench Verified, approaching Claude Opus territory. It runs natively on Chinese hardware at half the deployment cost. Kimi K2.5 from Moonshot AI adds another contender with strong agentic capabilities. What this means for professionals: the cost of AI mastery is dropping toward zero. The models are free. The compute is getting cheaper every quarter. The only remaining bottleneck is knowing how to prompt them effectively.

🚨BREAKING: Alibaba tested AI coding agents on 100 real codebases, spanning 233 days each. the agents failed spectacularly. turns out passing tests once is easy. maintaining code for 8 months without breaking everything is where AI collapses. SWE-CI is the first benchmark that measures long-term code maintenance instead of one-shot bug fixes. each task tracks 71 consecutive commits of real evolution. 75% of AI models break previously working code during maintenance. only Claude Opus 4 stays above 50% zero-regression rate. every other model accumulates technical debt that compounds over iterations. here's the brutal part: - HumanEval and SWE-bench measure "does it work right now" - SWE-CI measures "does it still work after 6 months of changes" agents optimized for snapshot testing write brittle code that passes tests today but becomes unmaintainable tomorrow. Alibaba built EvoScore to weight later iterations heavier than early ones. agents that sacrifice code quality for quick wins get punished when consequences compound. the AI coding narrative just got more honest: most models can write code. almost none can maintain it.

@RailMinIndia @RailwaySeva These are my details, after boarding I found a used diaper below the seat. Please find out the passengers travelling before me and hold them accountable for littering into public property. Civic sense will develop only when the enforcement is strict.

@Motion_Viz SYSTEM

2 months ago: Abandoning projects by week 4. Today: Zero abandoned projects. The shift? Stopped re-planning. Stopped re-researching. Stopped re-debating decisions I already made. The system: - Claude thinks through problems - NotebookLM remembers every decision - Markdown tracks what ships When someone joins my project now: → Point them to the .md files → Point them to the NotebookLM notebook → They're up to speed in 1 hour, not 1 week. Reply "SYSTEM" and I'll DM you the exact workflow.

3.5 years ago I transitioned from being a software engineer to running a mobile gaming company This is what I wish I knew at the beginning 1. Your players will call you greedy or your company will go out of business. 2. Ads work. As a software engineer, I thought no one clicked on them. I was wrong. 3. Organic content is not king. It’s spray and pray. Social media ads are like a sniper for finding customers. Both are good. 4. You need to earn the right to make a big game. Start with a game that you can make in less than 6 months. 5. You can’t MVP a game. Wait until you have a tight polished core loop to launch. 6. You can’t “fix” a bad game. Make a new one. 7. To make a great game you need someone that is 10x at: game design, coding, and visual design. This is usually 3 people. Partner/hire what you are missing.

Gemini 3.0 + Lindy + Perplexity = AI Content Infrastructure that generated 30M views last quarter... This 3-agent system replaces entire content teams automatically using AI strategist + producer + analyst architecture... → No more $15K-$30K monthly payroll for 4-person content teams → No more 20+ hours weekly spent planning content calendars manually → No more creative bottlenecks killing your posting velocity → No more analysts tracking metrics in 10 different spreadsheets Just 3 AI agents → autonomous content infrastructure that runs 24/7. Here's how it works: → Strategy Agent (monitors trends, identifies angles, builds calendars automatically) → Production Agent (generates platform-native posts, maintains brand voice across 1000+ posts) → Analysis Agent (tracks engagement, identifies patterns, optimizes continuously) → Multi-Platform Publishing (LinkedIn + Twitter content deployed simultaneously) → Performance Loop System (learns what works, compounds results weekly) Built with Fortune 500 content velocity. Runs 24/7 without creative bottlenecks. Zero payroll overhead. Enterprise quality. Results from deployments: • 30M+ organic views generated • $500K+ in qualified pipeline revenue • 25 posts weekly (up from 5 posts with manual teams) • One creator: $20K writer team → $500 AI infrastructure Want the complete system? Like + comment "LINDY" + repost, and I'll DM it to you. (must be following)

@aryanXmahajan LINDY

One thing you realise when you move abroad is how precious it is to simply breathe. When people ask me what my favourite thing about living in London is, I tell them “clean air.” And I can always see the surprise on their face. But if you’ve lived in Delhi, you know exactly why. Every winter feels like survival mode. You can’t see the sky, your chest feels heavy, and you start questioning how this became normal. I’ve moved my parents out too. Because at some point, it stops being about lifestyle. It’s about staying alive. If you can afford to, please move. You can rebuild everything except your health.

@tomcrawshaw01 SKILL

Writing viral posts will never be a struggle again... I packaged every context file, example, and framework that got me 4.5M+ impressions into one Claude skill. Just upload it to your project and Claude instantly references: - 20+ tweets with 100K+ views each - Proven hook formulas - Writing principles that actually work - Real feedback on what converts (and what flops) Most people prompt Claude with zero context and wonder why their posts sound generic. They're basically asking it to write blind. This skill changes that completely. You upload it once to your Claude account and it becomes part of Claude's memory. Now when you say "write a post about X," Claude pulls from battle-tested patterns. - It knows which hooks are overused and market-fatigued. - It understands how to structure posts for maximum engagement. - It references actual examples that crushed it. The skill teaches Claude to write posts that actually convert, and it never runs out of fresh angles. Claude gets better the more you use it together, because you can keep adding to the skill over time. This is the new way to add context to AI. No more copy-pasting examples into every chat. No more re-explaining your style from scratch. Follow + comment "SKILL" and I'll DM you the file

@Ai_here202 Resume

This Resume has an ATS score of more than 92🤯 This Resume helped many in getting an interview calls from companies like Google, Microsoft, Amazon, and many more. I have personally used this single-column resume in my job hunting and got amazing results I am sharing the exact similar editable ATS Friendly Resumes templates! To get it: 1. Follow me (So that I can DM) 2. Like & Repost 3. Reply "Resume" Follow me so I will dm immediately 💯

@elonmusk @xai Hi @elonmusk, instead of Grokipedia, why don't you name it XPedia? That kind of resonates with everything "X" in your products & companies - Space"X", "x"AI, "X".

We are building Grokipedia @xAI. Will be a massive improvement over Wikipedia. Frankly, it is a necessary step towards the xAI goal of understanding the Universe.

Excited to see @code (together with fast-agent) as one of the few MCP clients that fully implement the entire MCP spec! 🙌

🎉 @VisualStudio 2026 Insiders is here – and it marks one of the most ambitious steps forward we’ve taken with the IDE. 📖 devblogs.microsoft.com/visualstudio/v… 🆕AI woven directly into the developer workflow, performance improvements that reset expectations for speed at enterprise scale, and a modern design. We are just getting started and excited to be on this journey. Try it out and share your feedback to help share the evolution! #visualstudio

@vtchakarova @grok is this true?

🚨 According to the German FAZ, Trump has tried to reach Modi four times in recent weeks - all calls were refused. 🇮🇳 India’s official line continues to deny such contacts. 🇨🇳🇷🇺 Modi turns to the DragonBear avoiding Trump’s trap. One World, Two Systems - Delhi chooses hedging.

The Eisenhower Matrix is a powerful framework! It helped me stay clear and intentional, my whole life. ✅ Urgent & Important → I do it now 📅 Important, Not Urgent → I schedule it 🔁 Urgent, Not Important → I delegate it ❌ Neither → I eliminate it Simple and efficient.

@Lenskart_com I have sent you these details. However, I have no idea what do you mean by "temple" image here

@sumitparak Hi Sumit! We're really sorry for the experience you've gone through. Please share your registered contact details and the concerned order ID via DM along with 3-4 images of the product and a temple image so that we can get this checked and assist you. Thanks- Iti. twitter.com/messages/compo…

@Lenskart_com I ordered 2 frames with owndays lens (under buy 1 get 1 offer) and the 2nd frame was defective. No proper fitting and completely mis-aligned. I exchanged that for the same frame but again got the defective piece. Is this how you treat your customer?

Now i will have to AGAIN place an exchange order unfortunately. Very pathetic service. Is this the kind of product you deliver when a customer places an order under buy 1 get 1 scheme???