RT @ValDotTown: “99% of non-technical users have no f***ing idea how to deploy apps”
[until they use Claude Code + the Val Town MCP server…
Notes
Reposted by Steve
Text may be partial · Read original
- @TobinSouth yes! dm'd!
- @MaksimXBT lol the "limited" was referring to the fact that mcp apps don't allow any external resources, like tracking scripts!
- video at 1x speed. would y'all have preferred 2x?
- we shipped an "mcp app" so tool calls render with proper uis. pretty neat! most importantly, this means diffs render properly (not just as an unreadable json blob) also has (limited) html previews, logs, and more! https://x.com/stevekrouse/status/2061465229238616329/video/1
- @chrislakin hahahaha
- @thejasonchan ha! we're too late https://x.com/stevekrouse/status/2060474650405011691?s=20
- someone with a real fax machine got a fax an ai agent send and paid for with crypto! https://x.com/mj41Fantastican/status/2060471662139650084
- @leftcurvecletus There are two already! http://robotpostal.com http://postbridge.ai
- @shafu0x let's lock all the candidates in a room together until they pick who should get the job via smoke signal
- now you can all watch me get rich realtime https://x.com/stevekrouse/status/2060447673094729920 https://x.com/stevekrouse/status/2060468744950878706/photo/1
- @shafu0x @tempo very slick! got it working super quickly! poncho is the slickest agent with a wallet in the browser i've ever seen by far! https://x.com/stevekrouse/status/2060453449246843111/photo/1
- @CryptoCow33 @tempo not yet!
- my wife is belittling my new fax business. please more faxes so i can show her that i'm a real businessman https://x.com/stevekrouse/status/2060436594583076978
- @VicVijayakumar @aliceisplaying hahahahaha
- i made my first 20 cents from an api payment!!! what a rush!! https://x.com/stevekrouse/status/2060410355679961299 https://x.com/stevekrouse/status/2060436594583076978/photo/1
- @mj41Fantastican @tempo lol, shot you a dm
- @tanishqk @ValDotTown @tempo https://x.com/stevekrouse/status/2060429999388143851/photo/1
- @utekkare @tempo 🤣 lmk if you use it!
- @mj41Fantastican @tempo what's your fax number?
- @tempo https://agentfax.val.run
- @msg @tempo https://agentfax.val.run/
- I made a fax machine for your agents! ~~ Introducing agentfax ~~ Why? This morning, I had to fax the IRS. Claude Code got me 90% of the way there – even filling out the PDF! – but it couldn't actually send the fax So I had Claude Code find a fax API, wrap it in a @tempo proxy, and deployed on Val Town. Took me two hours end-to-end (while working on other stuff) I wonder if anyone will use it! Does anyone fax anymore?
- @KeshavY93629914 All the code!
- RT @CharlieMolthrop: Early numbers for Opus 4.8 look good. But can it pass the ultimate assistant benchmark? Meryl's opening rant from T…
Reposted by Steve
Text may be partial · Read original
- RT @kevinsuh34: been playing around with @ValDotTown - think of it as the fastest way to host a full-stack app, no local dev / deploy requi…
Reposted by Steve
Text may be partial · Read original
- Opus 4.8 live in Townie https://x.com/stevekrouse/status/2060111986998837701/photo/1
- my daily driver for val town has been claude code we just shipped a claude plugin which installs both: 1. our mcp server 2. skills for idiomatic val town patterns this is an improvement over the bare mcp server because it'll incrementally load the knowledge it needs
- https://github.com/val-town/plugins
- Why don't LLMs run their code more often? One of the first things I learned as an engineer was to run my code as early and often as possible: Step 1. Get *something* running. Step 2. Get it deployed. Step 3. Make a change. Confirm it runs. Repeat. Step 4. Send a PR / Deploy. At every point, your code runs or you get fast feedback on what broke it. LLMs – even in premier harnesses – often do the novice thing of writing thousands of lines of code before running & deploying. Sometimes they'll even confidently tell you they're done without ever testing any running code! Why? LLMs are writing professional code, but iterating like a freshman in CS 101. My guess is that they avoid running code, because it's so expensive relative to just outputting more code. In order to run code, an LLM has to do another turn through its tool loop. I think this is why agents tend to write a lot, then verify once at the end – or never. But for us humans, we've taken pains to optimize the edit + run flow into a single command, like cmd+enter, or just hitting save and hot reloading. This way, verification is free, instant, and woven into the loop. Why shouldn't it be the same for LLMs – a single tool call that makes the edit *and* runs it? I was surprised I couldn't find more people talking about this pattern. Aider, Cursor, Cline, Replit, Copilot all have "auto run after edit," but at the harness level: the LLM still issues a separate tool call and still pays the round trip, which is expensive and slow. Today, we shipped a small tweak to Val Town's MCP which should fix this. Now our edit file tools take an optional `run` object alongside any edit, so the LLM can test the change immediately after it's applied: { "tool": "replace_in_file", "input": { "val": "stevekrouse/my-api", "path": "main.ts", "old_str": "return 'hello'", "new_str": "return 'world'", "run": { "kind": "fetch_val_endpoint", "method": "GET" , "path": "main.ts" } } } I think this works for Val Town particularly well, because our deploys are 100ms, but I'm still confused why I haven't seen more of this pattern elsewhere. If you've seen prior art on fused edit+run tools, please send! The main alternative I considered is Cloudflare's code mode, which would be more composable and elegant. Instead of manually combining two tools like this, code mode lets LLMs compose tools itself in TypeScript. However, I decided to go ahead with this little experiment despite the "smell" because I think edit+run is a fundamental enough pattern to enable fast feedback loops that it was worth it
- I made a new side-project: an app to securely share secrets Usually, there's a tradeoff between security and usability For max security, you can use an oss tool like magic-wormhole, but usability suffers: you + your friend need to install it and use via terminal For max usability, you could use a web-based app, but that requires you trust it without being able to audit the code My app runs on val town, which means you can audit the running code If you'll pardon the blockchain analogy, val town lets apps function like smart contracts: you can see and audit the running code
- https://e2e.val.run/
- @CharlieMolthrop @OpenRouter @ElevenLabs @ValDotTown very entertaining
- RT @CharlieMolthrop: I hosted an open mic for ten AIs using @OpenRouter, @ElevenLabs, and @ValDotTown https://x.com/CharlieMolthrop/status/2059772171354882284/video/1
Reposted by Steve
- RT @CharlieMolthrop: @adelwu_ Couldn't help myself! http://coolworkbags.val.run Stack: - Codex - Anthropic Frontend Design Skill - Deployed vi…
Reposted by Steve
Text may be partial · Read original
- @bradnoble @EBogomolovs recorded for you @bradnoble https://x.com/stevekrouse/status/2059667390334468205?s=20
- Can we integrate 40 APIs in 20 minutes? https://x.com/i/broadcasts/1pKkOORPwwoKj
- i reorganized val town's templates http://val.town/explore/use-cases https://x.com/stevekrouse/status/2059665381602525475/photo/1
- @bradnoble @EBogomolovs Ok!
- @EBogomolovs Agreed 100%! 4 different vendors in 10 min shows how fast you can prototype now Before stripe projects or x402, it'd be at least 10+ min per vendor to get API keys With agents able to pay, vibe coding for prototyping and little apps is so powerful







