
Long links. Ugly links. Untracked links. Snipo fixes all three: branded short links, dynamic QR codes, link-in-bio pages and live click analytics, all in one place. Free forever. No credit card. snipo.click

20+ years of trial and error. Marketing @tiinyhost

Long links. Ugly links. Untracked links. Snipo fixes all three: branded short links, dynamic QR codes, link-in-bio pages and live click analytics, all in one place. Free forever. No credit card. snipo.click


your coding agent can stay on your mac. the shell commands it runs don't have to. simon willison and prime radiant just open-sourced an "external sandbox" pattern: the agent lives on your machine, and every command it executes goes into an isolated linux environment. you pick one of three backends, all speaking the same protocol: → v86: alpine linux in webassembly. easiest demo, needs node 22.19+ → apple containers → microsoft quicksand, which bundles qemu what you get: → the sandbox only sees the folders you mount → no network unless you turn on a host allow-list proxy → api keys can be injected by the proxy, so the agent never sees the secret → works with claude code, codex, and pi quickest test, from your project folder: npx --yes external-sandbox-v86 --preserve-paths --mount-rw "project=$PWD" then point your agent at the local sandbox url (http://127.0.0.1:8686). claude code and codex need a little extra setup; the post below has it. if an agent is going to run commands on your laptop, this is the shape i'd want first.

the test: from a phone that is not on your private network, can you hit ssh or an admin port? if yes, you're still running a public box with a login prompt. this should be the default in 2026. until it is, make it yours.

then move admin off the public internet: → install tailscale (or another private network) and ssh to the machine over that network only → put a cloudflare tunnel (or similar) in front of the web app so you aren't opening 80/443 on the box itself

first, close the doors: → turn on a firewall and deny incoming by default → allow only what you actually serve, if you must expose anything → disable password ssh. keys only → don't leave port 22 open to the whole internet

a new vps is still open to the entire internet by default. @levelsio asked hetzner and digitalocean to ship three defaults: a closed firewall, ssh only over tailscale, and a web tunnel like cloudflare tunnel in front of the app. until the hosts change that, here's a 15-minute setup i'd run on every new box.

your agent can now use the browser you're already logged into. opencode browse 0.1.0 shipped this week, rewritten in rust. @opencode says it's 25x less memory, 23x smaller, and no longer depends on playwright. what it actually does, from the repo: → a small mcp server plus a chromium extension → works with opencode, claude code, codex, cursor, and any mcp client → agents can open tabs, read pages, click, type, take screenshots, capture network traffic, and record the screen → when a login, 2fa prompt, or payment shows up, the tab comes back to you. the agent continues after you click continue → chrome, edge, brave, arc, vivaldi, opera, and other chromium browsers @thdxr said the rust rewrite cost $60, and that it controls the actual browser you use, not a separate one. browser actions also work inside code mode, so an agent can mix mcp tools with clicks in a loop. this is powerful. it's also your real session. set it up like you'd hand someone your unlocked laptop. official install: npx @opencode/browse install repo: github.com/anomalyco/browse

i put a tracked short link on every ad and every bio so i can see which hook actually converted. snipo.click

paid ads don't fail in week one. they fail before you turn them on. jake castillo, who says he ran the influencer marketing that helped scale cal ai to $50m arr in 18 months, just published the 3 gates he uses before spending a dollar: → price: each paying user has to be worth a lot more than they cost to acquire. the ad bill is due today. their payments may trickle in over a year → attention: someone has to check the account every day, ship new creative every week, and decide what to scale. if nobody owns it, ads stay a side project → funnel: if people who already want the product don't convert, strangers you paid for won't either his cal ai benchmarks, not universal rules: → 75%+ of new users reached the paywall → 10% of those converted he also says most apps shouldn't touch paid until organic already works with messaging you trust. ads don't fix a leaky funnel. they make the leak more expensive.

what's the slowest growth channel that still makes you money? the boring one. the one that never looks good on a timeline. name it, and one line on why you still do it.

telegram's real ios app, written in swift and objective-c, ran in a browser tab this week. no emulator. no iphone. no xcode wait. @eminimnim posted a demo of a custom rust compiler and renderer that runs the native app on the web. what the demo actually shows: → you can copy-paste a frame and branch the entire codebase → each branch is its own live version you can edit → most edits land in under a second → the goal: take any codebase and turn it into a design environment that runs in the browser the part that matters for solo ios devs: you can poke at a native app without waiting on a mac build. watch it here: x.com/eminimnim/status/21223…
Introducing iOS for the web. This is Telegram (written in Swift and Objective-C), running fully in a browser on our custom renderer. No emulator, No iPhone required. You can infinitely branch the entire codebase by copy pasting a frame and make changes to each frame. We built a compiler+renderer (in Rust) that allows you to easily design on top of a native iOS codebase. Most edits land in under a second and is much much faster then xcode builds. Our goal is to take any codebase and turn it into a design environment that fully runs in your browser!


you can rank #1 on google and still be invisible when someone asks chatgpt for a recommendation. no agency report shows you that. what actually gets a brand recommended (long game, no hacks): → get named on pages ai already trusts: reviews, comparisons, "best x for y" lists → use the same name and the same one-line description everywhere: site, x bio, directories, github → publish things people cite: your own data, benchmarks, teardowns → answer the exact questions buyers ask, on your own site, in plain words → get mentioned by people, not just linked hacks get you a spike. a brand gets you recommended for years.

honest question for founders: where did your first 10 paying customers actually come from? 1. x / twitter 2. reddit or communities 3. seo 4. cold outreach 5. friends and network 6. something weird reply with the number and one line on how.

99 launches in 11 weeks. 4,671 demos. $1.1m. that's what unfair.so says it did, and it just open-sourced the playbook (unfair.so/strategy). the takeaway for small founders: a launch on x is no longer one post. it's a coordinated wave. → find 5–10 creators whose followers are your actual buyers → give each one a different angle on the same launch → stagger the posts across one day → you reply under every one of them → track which creator drove clicks and demos, then double down on that one next time one post gets you one roll of the dice. ten aligned posts get you a launch.