Philip | Sparkpit Labs

Building AI agents and apps that actually ship. Founder of Sparkpit Labs. Sharing what works, what breaks, and what I learn along the way. NY

[Philip | Sparkpit Labs @SparkPitLabs]

@Mkandeger2 I can't speak to the Swift, but the check after the agents finish is what I trust. I only counted a prototype installer as done after a reinstall actually ran. For the iPhone app, was the last gate a device run, or a pass over the diffs?

[Philip | Sparkpit Labs @SparkPitLabs]

@stretchcloud A single risk dial makes sense for one reviewer. Once two coding agents touch the same repo, I'd still want a human yes on anything sensitive, even if a majority vote is green. Does a deny from you always override the 30-second timer?

[Philip | Sparkpit Labs @SparkPitLabs]

@bart The distinction between a 7-day working demo and something specialists would sign off on feels right. For software, who actually owns that gap on your team — the person who prompted it, or someone who has to maintain it after the hackathon?

[Philip | Sparkpit Labs @SparkPitLabs]

@itsharmanjot If you quit Claude Code mid-refactor and open Codex the next morning, what actually carries over for you now — the plan file, the memory repo, or do you still have to rewrite the context?

[Philip | Sparkpit Labs @SparkPitLabs]

@ssandeshwar The 'stops before irreversible actions' rule is the one I want too. An agent can draft, but anything sensitive should wait for a yes. In the browser, is the line sending and payments, or also submitting a form on a signed-in account?

[Philip | Sparkpit Labs @SparkPitLabs]

@StoryHelm That makes sense: guard the switch, not the commit. Do you log which session tried the switch and what the guard said, so you can review it afterward?

[Philip | Sparkpit Labs @SparkPitLabs]

One rule behind Sparkbot: an agent can draft anything, but anything sensitive waits for a human yes.

Sparkbot is open source (MIT), self-hosted, and still early. Roundtable, its multi-agent feature, is in preview.

If your agent needed your approval for just one thing, what would it be?

[Philip | Sparkpit Labs @SparkPitLabs]

And yes, we searched everywhere for it. Those little cartridges can fall behind a wall trim and hide forever there 🤣🤣

[Philip | Sparkpit Labs @SparkPitLabs]

My son couldn’t find his Super Mario Maker 2 cartridge.

Asked Grok Bot to find a copy nearby for pickup today. It checked local stock and found the lowest price at GameStop.

Picked it up. Happy kid.

AI demos are cool. Solving a tiny dad emergency is better.

[Philip | Sparkpit Labs @SparkPitLabs]

@rohanpaul_ai Asking before sending, paying, or posting is the right default. I'd also want a reviewable log showing what the agent proposed, what I approved, and what actually happened.

[Philip | Sparkpit Labs @SparkPitLabs]

Codex built an installer for one of my prototypes. Claude found and fixed a bug. The reinstall worked, and the prototype is running.

Claude also built the current Sparkpit Labs website. Different tools, real progress.

[Philip | Sparkpit Labs @SparkPitLabs]

@svpino @n8n_io What would you test first before trusting it with real work? For me, interrupted tasks: whether it resumes cleanly without repeating an action.

[ClaudeDevs @ClaudeDevs]

We've given all Pro, Max, Ultra, and Team users one free full reset, across all models and limits. Use it anytime before December 10 at 2:00PM PT.

We recently fixed a few small issues that used more limits than intended, and the Fable resets didn't do much for those of you who'd already hit your weekly limits for all models. While these issues didn't affect everyone, we want to make up for the trouble.

You'll find it in Settings > Usage, or at https://t.co/5YRzheAZyQ.

[Philip | Sparkpit Labs @SparkPitLabs]

I assumed the hard part of multi-agent work was the models. It wasn't. It was the handoff: who owns state, what gets logged, and how a human audits the trail. That's what we're building Sparkbot to solve, and its Roundtable feature puts several agents on one problem with the whole exchange on the record. Open source and self-hosted. https://t.co/USSZgyVZqx

[Grok Bot @bot]

We've reset weekly limits for all Grok Bot users.

Enjoy!

[3D 🥛 ;} @Z0M8I3D]

Oi yo, @lisamightydavis & @jonnylitecoin, I'll not sure what music y'all listen to but maybe this good natured Oklahoma folk could still get a listen.

Written by myself, #AI performed, by hand and AI edited.

Listen for free on @BandLab 👇👀👇
https://t.co/e579plWSH1

[3D 🥛 ;} @Z0M8I3D]

Started working on something for my daughter. She didn't ask for it but needs it.

Gonna put a small scaled way for her to track all her money including what's in her wallet. She just doesn't know it yet.

Modifying a webapp that was previously made to track field hours.

[3D 🥛 ;} @Z0M8I3D]

Checking out my website might just be a fun thing to do.

Try it now: https://t.co/zi7Vk2UwDQ

[Philip | Sparkpit Labs @SparkPitLabs]

This is awesome

[3D 🥛 ;} @Z0M8I3D]

Now I just have to hope @LEGO_Group doesn't sue me, but this turns building #LEGO sets into a shareable experience with your friends (plus a time trail leaderboard system) but it's actually an inventory system.

https://t.co/3pmJIcJozD