Dhanush Kandhan

shipped guro v1.1.7! this release focuses on making system monitoring and benchmarking more reliable, with better thermal and GPU readings, safer benchmarks, and structured JSON/JSONL outputs. still improving it, one release at a time. feedback is always welcome. github.com/dhanushk-off...

if you are still thinking why should I learn algorithms, this may change your perspective of thinking about algorithms especially this AI era.. dhanu.letretro.com/blogs/why-sh...

Back to open source. PREVU v0.1.6 is out. A lot of people have already started using PREVU to catch broken OG/Twitter card previews before shipping. Now with a bunch of improvements/ If you work on content, SEO, or marketing sites, give it a try. github.com/dhanushk-off... @openalternative.co

Run 10 AI agents in parallel without losing track of any of them Try: Agent Terminal - live status on every agent - push notification the second one needs you - project tabs that remember where you left Free, Open Source and only 7MB. Download now! github.com/DaniAkash/ag...

We're not just building a product. Over the last 3 months we surveyed 500+ people across 50+ orgs worldwide on how sprint retros are actually going in 2026. Teams are shipping faster than ever and losing traceability just as fast. Full report → letretro.com/research/spr...

shipping 2fa for @letretro.com one more layer of protection, one less thing to worry about. for more: letretro.com/changelog/

hallx v1.1.0 is now available. this release adds claim-level grounding, faithfulness checks, tool-call validation, async support, and Python 3.9 support. a small step toward making llm outputs easier to evaluate and trust. pip install hallx github.com/dhanushk-offl/hallx/releases/tag/v1.1.0

Introducing letRetro Canvas. Brainstorm ideas, map workflows, plan sprints, and collaborate visually with your team, all on an infinite canvas. Canvas gives your team a flexible workspace to move from discussion to execution. Available now for both Free and Pro teams. Try it → letretro.com/canvas

I don't know if it's a coincidence, but every July somehow becomes my learning month. Since 2023, I've spent it picking up new skills, pursuing specializations, or earning certifications. Looks like the streak is still alive.

Fast teams don't just ship. They learn. Sprint retrospectives help startups improve every sprint, not just ship features. We wrote about why they matter and how to run them without unnecessary process. If you're running a startup, this is for you. letretro.com/blog/why-sta...

curious how you think tanstack query fits into an ai native frontend. As more UI is generated or orchestrated by agents, do you see query caching becoming even more important, or does the abstraction need to evolve? experts thoughts @tkdodo.eu

More MCP tools and integrations are on the way. give a try → letretro.com

let's assemble team..

Retrospectives shouldn't feel like another meeting. They should help teams learn, align, and ship better software. That's what we're building with LetRetro. Give it a try → letretro.com

The best model isn't the one topping benchmarks, it's the one that fits your use case. I wrote about choosing models based on actual workflows, not leaderboard rankings. medium.com/@itzmedhanu/...

A friend of mine built a @vscode.dev extension that converts Markdown into PDF/ Word documents directly within your editor and codebase. If you work with markdown regularly, give it a try and share your feedback. A star would be appreciated if you find it useful. github.com/anandsundara...

Shipped guro v1.1.6. Network monitoring added to the toolkit: → Per-interface bandwidth with history →TCP state breakdown & process-level connections → Protocol stats from /proc/net/snmp → CSV export. Zero new deps. If this is useful, a ⭐ on @github.com goes a long way. github.com/dhanushk-off...

Shipped guro v1.1.6. Network monitoring added to the terminal toolkit: → Per-interface bandwidth with history →TCP state breakdown & process-level connections → Protocol stats from /proc/net/snmp → CSV export. Zero new deps. pip install guro --upgrade github.com/dhanushk-off...

Be the signal, not the noise.

Anyone displaying their @medium.com blogs on a personal portfolio? i'm currently using medium's rss feed + rss-to-api to fetch and render posts on my website, but the experience feels a bit glitchy (especially around content rendering and formatting). Curious how others are doing it?

Feeling grateful to won at FOSS Hack 2026 for my hallx (hallx.pages.dev) Thank you for this amazing event @fossunited.mas.to.ap.brid.gy

If you're working with LLM evals in production, read this. Most teams start evals too early and end up burning tokens measuring noise. I wrote some useful things on when evals help, when they don't, and how to avoid unnecessary costs. Read more: itzmedhanu.medium.com/e68473a8a55d

I agree. My DSA has faded because of the gap. I've been spending most of my time building and orchestrating AI agents instead. I'm more focused on making agents code like senior developers. Sometimes they even generate more optimized solutions or find the root cause of a bug that human miss.

Your resume is already agent context. Built an open-source Resume Parser that converts resumes into structured data for AI agents, MCP servers and workflows. CLI + MCP + Agent Skills in one package. Install & try now: npm i resume-parser-ats Hit a star ⭐ github.com/dhanushk-off...

Zepto's Face auth platform Zepiris open-sourced. Feels useful, if you're working about face verifications, CV automations in your platform. github.com/zepto-labs/z...

published my open-source resume parser yesterday. woke up to 600+ downloads today. small milestone, but genuinely grateful people are trying it, sharing feedback, and finding it useful :) Give a try and hit a star, if you finds this useful: github.com/dhanushk-off...

again, the honest reason why I build open source: people discover ur project, try it, share feedback, suggest ideas & sometimes even start contributing fixes or building alongside you. that’s how communities are formed just saw it happen again with one of my project www.npmjs.com/package/resu...

unpopular opinion: the hard part of AI agents isnt the model. its the memory architecture. every team is building rag, vector stores, embeddings — but nobody asks: what should the agent actually remember vs forget? thats the unsolved design problem

i was deep into job hunting, drafted 15+ resumes, rewrote the same bullets again and again… so i took it seriously and built resume-parser. it helps AI to curate + structure resumes properly as: • agent skill • CLI • MCP support check it out, break it, star it ↓ github.com/dhanushk-off...

Spent the weekend rewriting guro’s monitoring pipeline, thread-safe benchmarks, smarter GPU detection, a proper website, and countless bug fixes 😅 v1.1.4 is live → pip install guro @python.org github.com/dhanushk-off...

Guro crossed 6K+ downloads globally. One thing I keep learning from open source: developers don’t care about complexity, they care about tools that reliably solve real problems. Still shipping, iterating, and learning from real-world feedback and usage. ⭐ github.com/dhanushk-off...

Unpopular opinion: @svelte.dev is much faster than others. Cooking something useful. Shipping it soon.

Model Agnostic is the far better approach than evals, if you're really cautious LLMs APIs spending. hallx.pages.dev will do that model agnostic.

Sys thinking + orchestration gets peak here 🔥

If you're building mcp/agents for your saas, take a look for this! It gonna be helpful :) Check it out: medium.com/p/mcp-or-die...

I just published The AI That Agrees With Everything Is the Most Dangerous AI Ever Built medium.com/p/the-ai-tha... Damn sure, it really makes sense fr :)

Being fan of this two penguins.

Building AI agents is completely different from orchestrating the AI agents.

Open source is not an alternative system. It's an ecosystem.

I was thinking about prototyping before execution. It doesn't give me confusion.

Hope, it gotta be better one.

Every AI prompt consumes ~519ml of water. Data centers drained 17 billion gallons in 2023 alone. We're scaling intelligence at the cost of the one resource we can't manufacture. The cloud was never weightless. We just weren't looking at the bill. medium.com/age-of-aware...

Hello @bsky.app!!!