Groww

I don't know if you guys have noticed this or not, but Google seems to have raised its bar significantly when it comes to judging page quality. Over the last couple of months months, many websites are seeing: • A sharp decline in indexed pages • Crawled – currently not indexed counts going up rapidly • More pages getting ignored despite being technically fine • Thin, repetitive, or low-value/efforts pages losing visibility faster than before Which suggests that Google's expectations around page quality have become much higher than before. A few things that seem to matter far more now: • Information gain and originality • Genuine topical depth • Stronger page-level quality signals • Clear differentiation from similar pages already available on the web The interesting part is that many affected sites still have good authority and crawl frequency. Google is crawling them...just choosing not to index them. Feels like indexing itself is becoming a stronger quality filter now, not just a discovery mechanism. Anyone else observing similar trends across projects? #seo #google #marketing

🎉 Just got my 2nd contribution merged into HolmesGPT, the CNCF Sandbox SRE Agent project! Now you can apply chart-wide labels and annotations across every K8s resource the chart renders(Deployments, Services, RBAC, HPAs, ConfigMaps, NetworkPolicies, and all MCP server addons) with a single values entry: commonLabels: team: platform cost-center: observability Why it helps: FinOps tagging, team ownership, and policy-as-code (Kyverno/OPA) all rely on consistent labels. Before this, you'd either patch resources post-install or fork the chart. Now it's one line in values.yaml . Thanks to the HolmesGPT maintainers for the review. PR: https://lnkd.in/gf7WjC6z #OpenSource #Kubernetes #Helm #SRE #HolmesGPT #CNCF

Maintaining an indexable debug log has reduced iterations with agents and 10x-ed code quality! Coding agents are bad at retaining debugging lessons unless you give them memory. They’ll fix a bug today, then walk into the same class of mistake again a week later when you're building a different app. I packaged the workflow I’ve been using to fix this into debug-log-skill. Open-sourcing it today: https://lnkd.in/grW4KuDv It gives agents and developers: 1. A structured, searchable DEBUG_LOG.md , 2. A habit of checking past failures before making edits, 3. A simple way to log root causes, and 4. Turn bug fixes into prevention rules for future sessions. Works well with Claude/Cursor/any agentic workflows where continuity matters. Just plug it into your repo and make codegen a lot more reliable!

A new seller onboarding on an ecom platform - navigating listing flows, payment setup, shipping configs for the first time. A junior trader learning Bloomberg Terminal. A 55-year-old parent trying to file taxes on a government portal. Complex products lose users at onboarding. Just built and open-sourced HANDY: not another chatbot but an AI assistant that watches what you're working in, understands the context, and guides you without you ever leaving the app. Handy can sit on top of any tool and be the guide that actually sees what the user sees in real time. What makes it a powerful software: →It points, not just tell: A companion cursor literally flies to the button you need. →Voice + Chat + Floating Widget: Speak when your hands are busy, type when you want to think. A tiny floating pill on your screen for one-tap access. Powered by Apple 's native STT (for speed) and Sarvam ’s TTS (for expression). →Contexual memory per app: Switch from Figma to Xcode to Slack - your conversation switches with you. Each app and website gets its own chat history. Come back to Figma? Pick up right where you left off. As of now, the last 10 conversations are appended to the prompt. →Smart web search: 37 hardcoded regex patterns run a <1ms heuristic check on every query — if it matches something like "latest version" or "npm install," a web search fires async in parallel before the model even decides. If the model also wants to search, results are already cached. If it doesn't need to search, there's zero overhead. The fast path stays fast; the slow path gets faster. → Website specific: Handy reads the actual URL via Accessibility APIs, maps it to the website (GitHub, Notion, LinkedIn) and treats each site as a separate context. → 100% local: API keys in your Mac's Keychain. Chat history as JSON on your machine. Nothing on anyone else's server. FULLY OPEN SOURCE. No waitlist, no premium tier, no gatekeeping. Star it, fork it, build it in XCode and let me know your thoughts! Inspired by the open source community (shoutout Farza Majeed ). Giving it back. Demo attached. GitHub in the comments. Assistance has never been this handy ✋🏻 #opensource #ai

Weekend hack with Opus 4.7. A CLI to generate perf reports(TUI/HTML) for any app installed on android device. This is my second claude assisted project in Python and I'm getting more comfortable with language I haven't written in a decade! Link: https://lnkd.in/gy3RUdGM

dev bandwidth is no longer just headcount × sprint velocity. It’s increasingly a function of how many people can effectively wield AI tools. Not equal yet.Probably multiplicative, expecting it to be quadratic if not exponential soon.

Seems like even Anthropic can't build a proper chat bot (or they don't care enough to build a good one)...

Support Engineers are often the silent heroes behind stable production systems. While users experience seamless applications, support teams are constantly working in the background to monitor, troubleshoot, and resolve issues. A typical day as a Support Engineer includes: • Monitoring production systems • Handling incidents and service requests • Analyzing logs and database queries • Collaborating with clients and internal teams • Ensuring SLA adherence This role demands a strong blend of technical expertise and communication skills. Support is not just about fixing issues — it’s about ensuring business continuity and delivering a reliable user experience. Every day in this role brings new challenges, deeper learning, and opportunities to grow 🚀 #SupportEngineer #ProductionSupport #TechCareers #SRE #Learning

The bottleneck on AI-assisted software engineering is not the model. It is how your org moves context around the model. Two engineers on the same team ask Claude the same architectural question an hour apart. They get different answers, not because the model is inconsistent, but because each one pasted a different subset of the same docs into the prompt. I wrote a proposal for an SDLC design that treats this as the core problem. The argument: Every AI output in an org is a function of three inputs and most orgs are getting all three wrong. - Knowledge is scattered across Confluence, Notion, and people's heads - Skill is locked inside senior engineers who cannot scale themselves - Expectation is sloppy PRDs and half-finished solution docs The post proposes a git-backed, LLM-agnostic system that assembles all three automatically: one command, full context, zero manual prompt assembly. And a compounding loop where every shipped ticket makes the knowledge base richer and the skills sharper. This is a proposal with zero miles on it. I have not run it in production. The post says where it is weakest and invites pushback. If you are thinking about how to make AI usage across an engineering org less of a per-person lottery, this is for you. https://lnkd.in/gmKVDWSM