Insights

Practical guides and insights for SMEs in Gujarat to grow with technology.

Browse all topics →•567 articles available

Search Articles

Popular Tags

Showing 567 of 567 articles

LinkedIn Post

It looks like Claude Code can now read AGENTS.md for project instructions.

It looks like Claude Code can now read AGENTS.md for project instructions. AGENTS.md is an open standard, so Anthropic are finally coming to the table here. However, it doesn't seem to load skills from an .agents folder just yet. My workaround for using Claude and Codex on the same project has always been to keep a CLAUDE.md file telling Claude to read AGENTS.md and look at the skills list before doing anything. So it seems I no longer need to do this! This is from v2.1.277, so you need to update to that for it to kick in. How are you keeping your instructions and skills in sync across coding agents?

View on LinkedIn ↗

LinkedIn Post

I asked over 1,000 AI agency owners how they're signing clients.

I asked over 1,000 AI agency owners how they're signing clients. The method with the highest conversion rate was warm network reactivation. Messaging people you've already worked with and pitching them on a new AI service. One member's warm contact told him in the first meeting "I'm happy to pay you to do stuff." Another signed a $5K deal from a friend who already trusted him. I think people skip their warm network because it feels too easy or too awkward, but the conversion rate is so much higher than cold outreach that it's kind of silly to not start there. Breakdown of all the methods working right now, with numbers from real people, in the carousel.

LinkedIn Post

7 automations worth stealing

7 automations worth stealing Set them once. Review the output later. Based on Kilo’s webhook automation article: code quality sweeps, prototype branches, stale branch cleanup, release prep, coverage analysis, first-time contributor support, and post-deploy smoke tests. #AIAutomation #DevOps #KiloCode #DeveloperTools #EngineeringWorkflow

5/18/2026

View on LinkedIn ↗

LinkedIn Post

Stop treating Claude like a basic chatbot. 🤖

Stop treating Claude like a basic chatbot. 🤖 This checklist turns it into an infrastructure. 🛠️ Most people open the tool, type a prompt, and get generic outputs. 📉 The issue is not the model, but the foundation setup. ⚙️ Watch full video here: https://lnkd.in/e5mifEHE You can turn the default setup into a power user system in a few hours. ⏱️ Follow the 72-step framework to configure everything from your output style to account management. 📋 Build persistent context so the system understands exactly who you are and what you are building. 📂 Connect your tools, such as your communication and storage, to achieve zero contact switching. 🔗 Group your tasks into skills so you can automate repeatable workflows. 🔁 Organize your workstreams into separate projects to prevent context bleeding. 🗂️ Ensure your security and compliance settings are production-ready. 🔒 Use advanced techniques like parallel sessions to reduce processing time and boost efficiency. ⚡ What is the first configuration change you are making today? 🤔 ♻️ Repost: Share this checklist to help your network transform their AI setup and work smarter. ➕ Follow: Deven Goratela [https://lnkd.in/dQwsb2jA) as the go-to authority for staying ahead in AI and automation. #Claude #ArtificialIntelligence #Automation #Productivity #TechTips

5/18/2026

View on LinkedIn ↗

LinkedIn Post

Claude Code skills that sell

Claude Code skills that sell The boring ones are the valuable ones. Based on breakdown of 100+ Claude Code skills, this carousel covers the practical stack: Skill Creator, Superpowers, GSD, Review, Context Mode, and ClaudeMem. #ClaudeCode #AITools #AIAutomation #DeveloperTools #AutomationBusiness

5/17/2026

View on LinkedIn ↗

LinkedIn Post

The AI Career Opportunity Nobody's Talking About in 2026

The AI Career Opportunity Nobody's Talking About in 2026 It feels like the main narrative right now on "how to win with AI" is to start an AI automation agency. That's one path. But it's not the only one, and a lot of people listening to that advice don't actually want to run sales or deal with the stress that running an agency brings. There's another move almost nobody is talking about. IBM just surveyed 2,000 CEOs at companies averaging $5.8B in revenue. A few stats stood out: → 76% already have or are hiring a Chief AI Officer (up from 26% two years ago) → Only 25% of employees actually use AI at work, but 86% have the skills → 57% of Chief AI Officers at major companies were promoted from inside That 61 point gap between "can use AI" and "actually using AI" is the whole opportunity. Every CEO is staring at it. Someone has to close it. That someone is who gets promoted next. So there are really two paths into this seat: 1️⃣ External. Run an agency, take freelance gigs, consult. Prove value, get pulled in-house. 2️⃣ Internal. Quietly become the most AI fluent person at your current company. Ship workflows nobody asked for. Be the obvious name when leadership needs an AI strategy. I just dropped a 19 minute YouTube video breaking down the IBM study, both paths, and how to figure out which one actually fits you. The biggest takeaway: You don't need to change your role. You need to change which version of your role you are. Pick your path on purpose, and play to what you already love. Link to the full video is in the comments 👇

"Code is cheap now. Software isn't." (and all of you vibe coders are still confusing the two.) Everyone's hyped about building apps with AI in hours. But they're building personal scratchpads and calling it a business. Here's the actual difference: AI unlocked the ability to build something for YOUR specific annoying problem. Right now. Then throw it away when you're done. That subscription tracker you need for exactly your budget setup? Build it in 5 minutes. Use it. Move on. That's genuinely useful. But that's not competing with real SaaS. That's competing with doing things manually. The vibe-coded app works fine with 1 user. Sometimes 10. Then a real customer shows up and things get weird: > Concurrent sessions break > Credentials stored in the wrong place > Third-party API changes its limits at 3am Nobody thought about any of that. Because vibe coding doesn't know that exists. Real software isn't expensive because developers are slow. It's expensive because complexity is real, invisible, and always shows up eventually. We're not in a golden age of SaaS. We're in the scratchpad era. Disposable tools are powerful. Build them. Use them. Throw them out. But if you're selling to actual customers, with real data, at any kind of scale? Code being cheap doesn't change what it costs to build something that holds up. Full article in comments - worth the read 👇

5/13/2026

View on LinkedIn ↗

2385122

LinkedIn Post

Most people open Claude Cowork and just... type.

Most people open Claude Cowork and just... type. They're missing 80% of what it can do. Here are 5 ways I use it to actually get work done — not just generate text: → Auto-organize files without touching a folder → Send emails without opening my inbox → Summarize any doc in seconds → Schedule tasks that run on their own → Turn raw data into a shareable report Swipe through 👉 Which one are you trying first? Drop it below 👇

5/12/2026

View on LinkedIn ↗

LinkedIn Post

Stop Wasting Time Retraining Claude 🧠

Stop Wasting Time Retraining Claude 🧠 Claude often feels like a blank slate every time you start a new session. 📉 Stop burning tokens on re-explaining context and get better results faster. ⚡️ The secret is setting up a tiered memory architecture for your projects. 🏗️ Start with a simple claude.md file. 📄 Keep it under 200 lines to ensure the AI actually follows your persistent rules. 🤖 Use automemory for preferences, but don't rely on it for critical system instructions. ⚠️ For larger codebases, transition to a markdown knowledge base or vector search to handle complexity. 🔍 The ultimate setup combines hooks, skills, and MCP to bridge the gap between sessions. 🌉 Stop treating your AI like it has a five-minute memory. ⏳ Watch full video here: https://lnkd.in/gHuiBbxF What memory hacks have you found to make Claude smarter? 💬 ♻️ Repost this to help your fellow developers stop wasting time on repetitive AI prompts. ➕ Follow Deven Goratela [https://lnkd.in/dQwsb2jA) for actionable insights on AI and automation. #AI #Coding #ClaudeAI #DeveloperProductivity #Automation #SoftwareEngineering

5/12/2026

View on LinkedIn ↗

LinkedIn Post

From AI Slop to Autonomous Agents

From AI Slop to Autonomous Agents Level 1: The AI Slop Trap Many creators rely on AI to generate basic posts, leading to repetitive phrases and generic "AI-isms." → How to level up: Recognize default AI writing styles, write more specific prompts, and always iterate on outputs Level 2: The Taste Injector Stop letting AI guess your voice. Create a dedicated voice and tone guideline document. → Feed 3-5 of your best posts into Claude Code so it learns YOUR tone Level 3: Strategy & Ideation Build a system to pull and synthesize data from your preferred knowledge sources. → Turn manual research into skills that automatically gather and analyze trending information Level 4: The Creative Director (Multimodal) Expand beyond text to generate consistent visuals and carousels while keeping your brand identity. → Create reusable JSON prompt templates and use reference images Level 5: Scale & Repurposing Take one primary asset and convert it into various formats for different platforms. → Build skills that adapt tone and style per platform Level 6: Automation Streamline your content pipeline with automated routines while implementing human-taste checkpoints. → Schedule daily tasks so you don't have to trigger everything manually Level 7: Autonomous agents Fully automated systems that generate content continuously. ⚠️ Proceed with caution to protect your brand. No shortcuts. Systematic growth. Every level builds on the previous. Follow @devengoratela for more AI content tips 💾 Save this carousel 📤 Share with someone who'd benefit Comment "levels" for a breakdown --- #ClaudeCode #AI #ContentCreation #SocialMedia #Marketing #Automation #AIContent #CreatorEconomy #DigitalMarketing #Growth

5/11/2026

View on LinkedIn ↗

LinkedIn Post

The philosophy difference that actually matters

The philosophy difference that actually matters Breakdown of two open-source AI agents with different approaches: "gateway around agent" vs "agent around gateway." Swipe through for the full comparison → #AI #OpenSource #DeveloperTools #Hermes #OpenClaw #CodingTools #TechTools #SoftwareDevelopment

Your AI-generated websites look like everyone else's. 🤖

Your AI-generated websites look like everyone else's. 🤖 It is what I call "AI Slop"—generic, dark, techy templates heavy on purple gradients. 🤢 The truth? AI doesn't lack taste; it lacks context. 🧠 To build premium front-end experiences, you must move through the 7 Levels of AI Web Design: Level 1: The Prompt Monkey – Basic text descriptions that lead to generic outputs. 🐒 Level 2: The Skill Injector – Injecting specialized design rules directly into the context window. 💉 Level 3: The Visual Director – Moving below the surface to extract actual code architecture from image references. 📸 Level 4: The Cloner – Using site teardowns to extract raw HTML/CSS/JS files for accurate templates. 🧬 Level 5: The Customizer – Sourcing premium blocks and original visual storytelling with tools like Midjourney and Kling. 🛠️ Level 6: The Visual Tinkerer – Engineering premium micro-interactions, typography weighting, and parallax motion. 🖱️ Level 7: The Architect – The frontier of custom WebGL, complex 3D shaders, and gamified web experiences. 🏛️ The elite build workflow is simple: Inject, Teardown, Swap, and Tinker. 🔄 AI code generation is only as powerful as your ability to articulate taste. 💎 Full Video here: https://lnkd.in/dEHtnc89 Are you still using basic prompts, or are you ready to start architecting elite experiences? Let’s discuss below. 👇 ♻️ Repost this to help your network move past AI slop and start building premium web experiences. ➕ Follow Deven Goratela https://lnkd.in/dVt7VtDu to stay ahead of the curve in AI, automation, and elite digital design. #AIWebDesign #ClaudeCode #FrontEndDevelopment #DigitalTransformation #WebDesignTrends2026 #PracticalAI #NoMoreAISlop

4/20/2026

View on LinkedIn ↗

LinkedIn Post

Your Second Brain, Now AI-Powered

Your Second Brain, Now AI-Powered Complete Karpathy-style Obsidian RAG setup in 3 steps Turn your entire Obsidian vault into an AI-queriable knowledge base. 100% private, zero cost, runs locally. Save this for your next setup session! #ObsidianRAG #PersonalKnowledgeManagement #AIProductivity #SecondBrain #NoteTaking

4/20/2026

View on LinkedIn ↗

LinkedIn Post

Most AI coding agents jump straight into the deep end 🌊

Most AI coding agents jump straight into the deep end 🌊 They write code without a spec, ignore tests, and deviate from your vision within minutes. That is how you end up with a mess of technical debt. Superpowers [https://lnkd.in/d3TVjHtt ] changes the entire software development workflow for your agents 🚀 It starts by stepping back. Instead of just writing code, it asks what you are truly trying to achieve. It teases out a spec you can actually digest. Only after you sign off on the design does it build an implementation plan. The system emphasizes: * True red/green TDD (Test Driven Development) ✅ * YAGNI—You Aren't Gonna Need It 🚫 * DRY—Don't Repeat Yourself 🔄 Once you say "go," it launches a subagent-driven development process. Claude can work autonomously for hours without deviating from your plan 🤖 It handles the engineering tasks, reviews the work, and moves forward automatically. Your coding agent finally has the judgment it was missing. If you are building in 2026, you cannot afford to waste time on "AI slop" code. Use a framework that forces your agents to think like senior engineers 🧠 Have you tried using subagents for your development workflow? Let me know your experience in the comments! 👇 Full Video here: https://lnkd.in/dz_k4mzr ♻️ Repost this to help your network build better software with AI. ➕ Follow Deven Goratela [https://lnkd.in/dQwsb2jA) for the latest insights on AI and automation. #AI #SoftwareDevelopment #GitHub #CodingAgents #Automation #TechInnovation

4/19/2026

View on LinkedIn ↗

LinkedIn Post

Install Your AI Skill Stack with One Command! 🚀

Install Your AI Skill Stack with One Command! 🚀 Autoskills lets you install your entire AI skill stack with a single command. It is an open-source tool designed to automatically bridge your codebase with your AI agents. The manual approach involves: * Reading endless documentation 📄 * Writing repetitive boilerplate system prompts ✍️ * Downloading individual skill files 📥 * Dealing with missing dependencies 🧩 The Autoskills approach offers zero configuration and automated tech detection. Simply execute one command at your project root and let the engine take over. There is no need to pass a list of frameworks or languages manually. The engine's "X-ray scanner" automatically detects your environment by reading configuration files and package manifests. It builds an exact profile of your project's architecture and installs perfectly matched AI agent skills. This injects domain-specific intelligence directly into your workspace. Autoskills understands modern stacks at scale, covering: * Languages and runtimes 💻 * Backends, APIs, and data storage 🗄️ * Cloud tooling and AI technologies ☁️ It even offers native integration for Cloud Code users, consolidating installed files into a single "claw.md" summary. By aligning your technology stack with pre-optimized intelligence, Autoskills transforms generic AI into your domain expert. Have you tried automating your AI workflow setup yet, or are you still writing prompts from scratch? 👇 Full Video Here: https://lnkd.in/guHgkxJg ♻️ Repost this if you're ready to 10x your developer productivity with automated AI agent skills. ➕ Follow Deven Goratela [https://lnkd.in/dQwsb2jA) for the latest tools and strategies to stay ahead in AI and automation. #Autoskills #AIDevelopment #OpenSource #SoftwareEngineering #DeveloperTools #AIWorkflow #Automation

4/18/2026

View on LinkedIn ↗

LinkedIn Post

Anthropic just made it possible to run AI agents in the cloud without building the infr...

Anthropic just made it possible to run AI agents in the cloud without building the infrastructure yourself. Managed Agents was launched on the 8th April, and it does something quietly significant... Until now, if you wanted an AI agent that actually runs reliably in production, you needed engineering. Not prompt engineering. Software engineering. Sandboxed environments, monitoring, permissions, execution loops. For most marketing teams, that was a wall. Managed Agents turns that wall into a subscription. Why should you care even if you're not building agents today? Because the tools you're paying for are being rebuilt with agent infrastructure at the core. Canva just acquired an AI agent management company. HubSpot is pricing AI agents at a dollar per qualified lead. Meta runs entire ad campaigns autonomously from a URL and a budget. The question isn't whether agents will affect your work. It's whether you'll understand what's happening when they do. One thing you can do right now: document your workflows clearly enough that you could hand them to a competent stranger with written instructions. If you can't do that for a human, you can't do it for an agent either.

Hire Your First AI Employee Today 🤖

Hire Your First AI Employee Today 🤖 Stop doing repetitive work manually. Claude Code 2.0 just introduced native scheduled tasks, effectively giving you a 24/7 AI employee that self-corrects and improves over time. Here is how to set it up in seconds.

This Claude Code plugin enables "agent swarms". Meaning you can manage 10+ sub-agents in one chat. But more than that, it instills a building methodology into Claude Code that has dramatically improved the quality of my builds. Here are the 5 steps broken down into skills: 1️⃣ Brainstorm Clarifies what to build before a line is written: - Scans your repo for existing patterns via a research agent - Proposes 2-3 approaches with tradeoffs considered - Writes an ideation doc with decisions captured 2️⃣ Plan Turns the ideation doc into a spec the agent can execute: - Parallel research agents scan your codebase and docs - An agent analyzes the concept for gaps and edge cases - Outputs a plan file with tasks, dependencies, and acceptance criteria 3️⃣ Work Builds out the plan with QA checks: - Swarm mode spins up parallel agents for independent tasks - Runs a "system-wide test check" after every change - Incremental commits only when tests pass 4️⃣ Review Code review to iron out issues before shipping: - Multiple agents review the output across all areas - Findings triaged with P1/P2/P3 - Agent confirms solution doesn't lie within previous context 5️⃣ Compound Captures every solved problem while the context is still fresh: - 5 sub-agents extract root cause, the fix, and prevention tactics - Saves it to a searchable knowledge base inside your repo - Future sessions search past fixes automatically before planning ----- The last step is the key. "Each unit of engineering work should make subsequent units easier, not harder." I'm not an engineer. But this workflow improves practically any build in Claude Code. And the "swarms"? Any step can spin up parallel sub-agents to divide the work. 10 independent tasks = 10 agents building simultaneously. Watching it happen makes you feel like an AI overlord. This is an open-source plugin comprised of 29 agents. Absolute game-changer, shout out Kieran Klaasen. Comment "repo" if you want me to send over the GitHub link.

How to run Claude Code like the engineer who built it. (Boris Cherny, the head of Claude Code, just revealed his exact workflow) Follow these 6 steps. ↓ Most engineers use Claude Code like a chatbot. They prompt one session and wait. You can run 10 to 15 sessions in parallel and ship code simultaneously instead. 1/ Set Up Parallel Sessions -Terminal: Run 5 Claude Code sessions simultaneously in your terminal -Web: Run 5 to 10 additional sessions in the browser at the same time -Combined: Every session works on a different task while you focus elsewhere 2/ Build Your CLAUDE.md File -Self-improvement loop: After every correction, Claude updates its own rules so the mistake never happens again -Lessons file: All rules live in tasks/lessons.md and are reviewed at the start of every session -The command: "After every correction, end with: Update your CLAUDE.md so you don't make that mistake again" 3/ Set Workflow Orchestration Rules -Plan Node Default: Enter plan mode for any task with 3 or more steps before writing a single line -Subagent Strategy: Offload research, exploration, and parallel analysis to subagents to keep main context clean -Verification Before Done: Never mark a task complete without proving it works first 4/ Structure Every Task in 3 Parts -Part 1: Write plan to tasks/todo.md with checkable items before starting -Part 2: Execute with high-level summary at each step, mark items complete as you go -Part 3: Add review section to tasks/todo.md and update tasks/lessons.md after corrections 5/ Set Core Principles -Simplicity First: Make every change as simple as possible, minimal code impact -No Laziness: Find root causes, no temporary fixes, senior developer standards -Minimal Impact: Changes should only touch what is necessary, never introduce bugs 6/ Delegate Entire Problem Types -SQL: Boris has not written a single line of SQL in 6 months, Claude pulls BigQuery data directly via CLI -Bug fixing: When given a bug report, Claude fixes it autonomously, zero context switching required -CI tests: Claude fixes failing tests without being told how This is just one way to use Claude Code the way its builder actually uses it. Want the full Claude Code system? → Comment "CODE" and I will send it to you

Stop writing bad prompts today. 🛑 Steal these 8 proven frameworks instead. 💡 8 ChatGPT Prompt Frameworks To Master AI 🧠💻 Most people treat AI like a simple search engine. 🤖 But if you want high-quality work, you need structure. 🏗️ Here are 8 prompt frameworks you can use right now to get better results. 🚀 1. R-T-F 📝 Set the Role, define the Task, and choose the Format. ✅ 2. S-O-L-V-E 🧩 Great for complex planning. Detail the Situation, Objective, Limitations, Vision, and Execution. 📈 3. T-A-G 🎯 Keep it simple. State the Task, the Action needed, and your Goal. 🏆 4. R-A-C-E 🏁 Perfect for business use cases. Give the Role, Action, Context, and Expectation. 💼 5. D-R-E-A-M 💭 A project manager favorite. Define, Research, Execute, Analyse, and Measure. 📊 6. P-A-C-T 🤝 Tackle tricky bottlenecks. State the Problem, Approach, Compromise, and Test. 🧪 7. C-A-R-E ❤️ Provide Context, Action, Result, and an Example. This sets crystal clear expectations. 🔍 8. R-I-S-E 🌅 Great for leadership tasks. Assign the Role, Input, Steps, and Expectation. 👔 My personal favorite for daily work is R-T-F because it is fast and highly effective. ⏱️ Which of these frameworks are you going to try on your next prompt? 🤔 Let me know in the comments. 👇 #ChatGPT #ProductivityTips #PromptEngineering #WorkSmarter #TechTrends #Efficiency #FutureOfWork

ChatGPT vs Grok vs Gemini vs Claude vs Perplexity — 5 leading AI tools explained. Best use cases, strengths & pro tips. No hype, just practical insights to choose the right tool. Save this post & read the full thread Credit: piyascode9 on Twitter/X

2/11/2026

View on LinkedIn ↗

103739173

LinkedIn Post

RAG isn't one algorithm anymore.

RAG isn't one algorithm anymore. Retrieval-Augmented Generation started simple. Retrieve documents, feed them to an LLM, generate an answer. That was 2020. Now RAG is a design space. Different architectures solve different problems. Accuracy, speed, reasoning depth, compliance. Here's what's actually being used: Core Architectures • Standard RAG: The original. Retrieve once, generate once. Still foundational but not enough for complex tasks. • Hybrid RAG: Combines keyword search with embeddings. Better recall. Now the default in production. • Recursive RAG: Multiple retrieval steps. Later retrievals depend on earlier reasoning. Needed for multi-fact queries. • Self-RAG: Model decides when to retrieve. Checks if the context is relevant. Reduces hallucinations through self-reflection. Knowledge-Structured Retrieval • Graph-Augmented RAG: Retrieval grounded in knowledge graphs. Enables entity linking and relationship reasoning. Common in regulated and enterprise domains. • Knowledge-Enhanced RAG: Adds ontologies or symbolic constraints. Improves consistency. Often overlaps with graph-based systems. Query and Retrieval Optimization • HyDE: Embeds a hypothetical answer instead of the query. Works better for short or vague questions. Widely adopted. • Query-Transforming RAG: Rewrites or decomposes the query before retrieval. Often yields bigger gains than changing embedding models. Memory and Temporal Context • Memory-Augmented RAG: External memory beyond the prompt window. Supports long-term context. Used in agentic systems. • Streaming RAG: Retrieval over continuously updating data. Logs, feeds, event streams. Requires time-windowed indexing. RAG today is a system design space. Not a single technique. Repost this if you found it useful. Save it for reference when building your next RAG system.

2/11/2026

View on LinkedIn ↗

132826195

LinkedIn Post

The OpenClaw hype is real.

The OpenClaw hype is real. But I’m ignoring it. ClawHub just dropped, and it’s essentially the App Store for AI. Anyone can build a skill, list it, and charge a subscription. The barrier to building software didn't just lower—it vanished. That’s a massive shift, but it’s also a massive distraction. Tech moves at light speed, but people don’t. Tens of millions of businesses still have no idea how to identify a use case, build a system, or get their teams to actually log in. That’s the "Rock vs. Sand" dilemma: ️ Building on Sand (The Marketplace) You build a cool micro-SaaS. You get some quick traction. Then, a platform update kills your tool or a giant copies it for free. You’re in a race to the bottom on code speed against the entire world. ‍♂️ Building on Rock (The Service Agency) You solve the "boring," expensive problems that keep CEOs awake. ☕ * Strategic consulting * Process auditing * Change management * Team training Companies are drowning in new tech. They don't need another subscription; they need a lifeline. The biggest opportunity right now isn't selling more shovels. It’s teaching people how to dig. ️ New platforms are tools, not business models. Use them to work faster, but don't let them become the mission. Are you chasing the marketplace trend, or doubling down on services? Let’s talk in the comments. #AI #OpenClaw #TechStrategy #BusinessGrowth #Consulting

2/10/2026

View on LinkedIn ↗

LinkedIn Post

In the bustling tech hubs of Bengaluru, a quiet revolution is revealed. Sarvam AI, an A...

In the bustling tech hubs of Bengaluru, a quiet revolution is revealed. Sarvam AI, an AI startup, has reportedly taken the global AI world by storm, signalling India’s growing ambition to build globally competitive sovereign artificial intelligence technologies and research ecosystems. With its latest models, Sarvam Vision and Bulbul V3, the company claims to have outperformed some of the biggest names in artificial intelligence, including Google Gemini and OpenAI’s ChatGPT. As per Pratyush Kumar, the vision of Sarvam AI officially released information on X, announcing a state-space-based 3-billion-parameter vision language model with the best results in digitisation in English and other Indian language. #sarvamai #chatgpt #gemini #aichat #makeinindia #corporatecurrent

2/10/2026

View on LinkedIn ↗

320104

LinkedIn Post

Google AI Subscriptions Now Include Cloud Credits! ☁️

Google AI Subscriptions Now Include Cloud Credits! ☁️ Building AI prototypes is easy. Deploying them is where it usually gets expensive. Google just changed the game for solo devs and small teams. As of late January, Google AI Pro and Ultra subscriptions now include monthly Google Cloud credits. Google has merged their AI subscriptions with the Google Developer Program, giving Pro and Ultra users monthly Cloud credits at no extra cost. Here is the breakdown of the new perks: * Google AI Pro: $10/month in Cloud credits. ⚡ * Google AI Ultra: $100/month in Cloud credits. This isn't just a random discount. It’s a direct bridge from "messing around" to "going live." You can now build and refine your prompts in Google AI Studio or the new Antigravity IDE, then push straight to production using your credits for: * Vertex AI deployment. * Cloud Run hosting. ☁️ * Gemini API calls beyond the free tier. For anyone building lightweight agents or personal automation tools, this effectively makes your hosting and API usage free. You are no longer managing two separate bills just to keep a small project running. ️ While high-traffic apps will still need a proper budget, this is plenty to get your MVP off the ground without hitting a paywall on day one. If you’re already a subscriber, you just need to head to the **Google Developer Program** dashboard to activate these benefits. ️ Is $10–$100 a month enough to cover your current side projects, or have you already scaled past that? Let me know what you’re building below! #GoogleCloud #GenerativeAI #GoogleGemini #SoftwareDevelopment #CloudComputing #TechNews #VertexAI #WebDevelopment #AIStudio #DevOps

2/10/2026

View on LinkedIn ↗

LinkedIn Post

85% of queries. One AI. ️

85% of queries. One AI. ️ Public service portals are often famous for long wait times and complex navigation. We decided to change that story. By deploying a specialized AI chatbot, we’ve successfully automated 85% of routine citizen inquiries. This isn't just about answering questions—it's about digital equity and operational speed. Here is the impact of moving from "search and find" to "ask and resolve": Instant Accessibility: Citizens get accurate answers 24/7 without waiting for office hours. Multilingual Support: The bot speaks over 75 languages, breaking down barriers for diverse communities. Staff Reclaimed: Teams are no longer buried under repetitive FAQs and can focus on complex cases that need a human touch. Frictionless UX: From permit status checks to document requests, the AI provides a direct path to the result. The "Digital Front Door" isn't a future concept—it’s the new standard for efficient governance. ️ By handling the "basics" at scale, we allow the human workforce to do what they do best: empathy-led problem-solving. Is your organization ready to automate the routine and prioritize the human?

2/10/2026

View on LinkedIn ↗

LinkedIn Post

Most people use Claude Code backwards.

Most people use Claude Code backwards. They throw a vague prompt like "build a login system" and then spend two hours in a "back-and-forth" death loop fixing bugs. There is a better way. Write the spec first. Before you touch the terminal, spend 20 minutes mapping out exactly what you need in a simple Markdown file. Think through: * Specific requirements ✅ * Edge cases and error handling ⚠️ * API structure and types ️ * Testing criteria Then, point Claude Code directly at that file: "Implement the auth service defined in auth-spec.md" The result is a total game-changer. Because the context is clear, Claude builds multiple files, handles the "forgotten" OAuth logic, and writes tests that actually pass the first time. Plus, that spec doesn't disappear into a Google Doc graveyard. It lives in your repo as permanent documentation for your future self. Stop guessing and start specifying. Your velocity will thank you. ⚡️ Have you tried "Spec-Driven Development" with AI yet, or are you still prompting on the fly? P.S.: Coming up tomorrow - The Perfect AI-Ready Spec Template

2/10/2026

View on LinkedIn ↗

LinkedIn Post

This meme is funny.

This meme is funny. And also… painfully accurate. npm install pip install “looks fine…” Agentic coding makes this way more dangerous. Because the agent doesn’t hesitate. It will happily pull in 5 new libraries to solve a tiny problem. Fast. And suddenly you’re not writing code anymore. You’re importing risk at scale. You can’t treat dependency hygiene as a “nice to have” anymore when coding at this scale and speed. From an ongoing agentic coding experiment at Nova Ventura #unglamorousrealityofAIadoption #vibecoding

Most founders don’t know they’re paying for 3+ tools that do the same job. And no one on the team tells them.Until one day, something breaks. Data’s i

Master the Tolerate-Invest-Migrate-Eliminate framework to make confident, data-driven decisions about every application in your technology landscape.

1/20/2025•1 min read

Read →

Application Portfolio RationalisationIT Strategy

Application Portfolio Rationalisation: The Complete Guide for SMEs

Learn how to evaluate, optimize, and streamline your application portfolio to reduce costs, eliminate redundancy, and align IT investments with business strategy.

1/15/2025•1 min read

Read →