varun

varun

@apvarun.com

Indiehacker - Engg. Lead from ๐Ÿ‡ณ๐Ÿ‡ฑ Currently https://sketchnote.app Curated product launches - https://launches.uicomet.com Gradients - https://gradients.party ๐Ÿ‘จโ€๐Ÿ’ป https://apvarun.com

991 followers509 following203 posts

pulseos.finance Local-first personal finance platform that runs transaction categorization and natural-language ledger queries on on-device LLMs orq.ai LLM gateway that enforces organization policies, token spend limits, MCP controls, and full audit logs before requests leave your perimeter

pulseos.financePulse Local: budgeting that never leaves your machineOther apps keep your data local and still send your question to someone elseโ€™s model. We run the models on your machine too.

Four infrastructure tools for moving AI agents into production archify.si Compiles codebases and system descriptions into interactive architecture maps directly inside coding agents mecatl.dev Cloud-native runtime for hosting agents on K8s with permission gates, sandboxed environments, and logs

archify.siArchify โ€” AI Agent Skill for Interactive DiagramsArchify is an MIT-licensed AI agent skill for validated HTML diagrams from descriptions or code, with interactive exploration and browser export.

A few newer builder tools I came across this week: roomote.dev a single-tenant cloud coding agent that works from your repo and returns PRs with live previews ledge.sh Markdown notes where code blocks run in place, including on an SSH server

roomote.devYour own cloud coding agent.Roomote is an AI engineering teammate that ships reviewable PRs โ€” hosted by us in about two minutes, or self-hosted on your own infra. Single-tenant either way, source-available, bring any model. Solo...

Building with agents? Save these: dev.fast AST-based code review, DB diagrams, and agent trajectory inspection skills.lassejlv.dk/skills/build-gโ€ฆ GPUI guidance for state, accessibility, tests, and packaging flask.do Turn spoken and drawn video feedback into timestamped agent comments โ€ฆ

dev.fastWhiteboard by /dev/fast โ€” The canvas for thoughtful softwareWhiteboard is an open-source canvas for thoughtful software development. Critique interactive visualizations, diffs, and agent trajectories, all from one interface.

- treg.to/ - one MCP key for 3,600+ paid data and automation endpoints, billed per call - ollaya.dev/ - open local decision models for calibrated yes/no, choice and scoring requests - github.com/remorses/gpuix - React bindings for Zed GPUI, for native apps without Electron

treg.totreg.to: OpenRouter for agent tools and data, pay per callOne MCP server, one key: 3,700+ tools across 99 providers for SEO, enrichment, ads, social and scraping data at the provider's own rate, no markup.

here are 4 builder tools that take a different route around the usual stack. save these ๐Ÿ‘‡ - devfra.me/ - define a devtool once, then mount it in any framework, a CLI, a static report, or an agent - github.com/lovablelabs/oj - Rust-native React builds aimed at lower memory use and faster cold start

devfra.meBuild a devtool once. Mount it anywhere.Devframe is a framework-neutral foundation for devtools. One definition becomes a Web Standard handler you can mount into any host framework, ship as a CLI or static report, and expose to coding agent...

- stalkr.ai / watch public X, Reddit, YouTube, and LinkedIn mentions, sort them into bugs, ideas, complaints, and testimonials - clarity.addy.ie / an open-source AI writing skill that interviews, rewrites, or reviews your product copy without flattening your voice what else helps you ship an app?

stalkr.aiStalkr โ€” Social listening for startupsMonitor X, Reddit, YouTube, and LinkedIn mentions for your startup. Alert your team or AI agents through email, webhooks, MCP, and API.

- is-agentic.com โ€” scan a public site for what AI agents can discover and use, with evidence for each gap - kolu.dev โ€” browser terminal workspace that keeps sessions alive and can add remote hosts over SSH what else belongs in this list?

kolu.devkolu โ€” the best way to run terminalskolu is a browser terminal workspace that spans machines: point it at a box over ssh and the whole canvas runs there โ€” terminals, code, git, agents โ€” backed by padi and kaval on every host.

The idea came from my own kitchen drawer full of expired warranties and receipts I couldn't find when I actually needed them. So I built the thing I wanted: scan once, search forever, get told before it expires. All on-device. That's Papertuck

varun@apvarun.com

Launched today ๐ŸŽ‰ Papertuck: your paperwork vault. Scan once, find it when life asks, get reminded before things expire. Local-first, no account, no ads apps.apple.com/nl/app/paper...

here are 6 UI resources worth saving ๐Ÿ‘‡ - great-ui.com โ€” 31 ready-to-use React and Tailwind components - github.com/nutlope/variate โ€” generate 4 code-based UI variations, then keep the one you want - tryanim8.com โ€” browser motion editor with editable keyframes and animated SVG export

- logoinspo.com โ€” 1,200+ logo references, filterable by style, industry, and color - appllama.io โ€” browse onboarding and paywall flows from 860+ top-earning iOS apps - goey-toast.vercel.app โ€” React toast notifications with morphing animations and promise tracking what else am I missing?

The illusion of choice in modern dev: - Terminal app - Agent harness - AI skill - Menubar app - Markdown app What are you shipping this week?

If I can only choose 1 skill to install, it would be this one from @cursor_ai > thermo-nuclear-code-quality-review The skill does a very good job at reviewing code across a bunch of parameters

varun reposted

It's been 3 years since | last updated my blog, but last weekend I finally sat down and wrote again. I put together some thoughts on low-context engineering: designing systems, tools, and workflows that are easier for humans and Al agents to understand, modify, and review safely.

It's been 3 years since | last updated my blog, but last weekend I finally sat down and wrote again. I put together some thoughts on low-context engineering: designing systems, tools, and workflows that are easier for humans and Al agents to understand, modify, and review safely.