I shipped a game in 4 hours.

Hey — and welcome to Sunday Signal.

Thursday evening I didn't have a new product idea. By Thursday night I had submitted one to the App Store.

It's called Spire. It's a one-tap stacking game. It went from "let me brainstorm something with an agent for 30 minutes" to "binary uploaded, App Store review pending" in roughly 4 hours.

Four hours. From nothing.

I'm going to tell you what I think actually happened, what it means for anyone building solo right now, and why I'm starting to think the word solo is doing some heavy lifting it shouldn't be doing anymore.

📡The Signal: 3 AI moves worth knowing this week

1. Anthropic just filed for IPO with a ~$47B run-rate

Last week Anthropic closed a $65B funding round at a $965B valuation. This week, Anthropic confidentially filed a draft S-1 registration with the SEC. As of May, the company's revenue run-rate is roughly $47 billion annualized — about 4x what their Q2 numbers projected just two weeks ago. The lab in most of our stacks is going public, and the growth curve is sharper than almost any private tech company before it.

Why it matters for you: When the AI lab in your stack is filing for IPO, two things shift. The product gets more enterprise-shaped (priorities, sales motion, support hierarchy). And every competitor of yours is reading the same prospectus. Use the leverage while it's still cheap, and start building habits that aren't tied to any one lab.

Sources: Reuters on Anthropic's S-1 filing · The Information

2. Google launched a 24/7 personal AI agent

Google introduced Gemini Spark, a cloud-based personal AI agent that keeps working in the background even when your phone is locked. Spark is pitched as an "active partner that does real work on your behalf" — Daily Briefs, calendar handling, ongoing tasks across your digital life. It rolls out to Google AI Ultra subscribers next week.

Why it matters for you: This is the first mainstream lab pitching an agent that runs for you while you sleep. Whether or not Spark turns out to be good, the framing is the future. Solo founders who already build their own agent stacks won't blink. Solo founders who haven't started will spend the rest of 2026 catching up.

Sources: Google announcement · TechCrunch on Gemini Spark

3. Google and Microsoft are coming for Anthropic's coding crown

Google announced an AI developer subscription at $100/month (less than half Claude Code's Max tier), bundled with Gemini 3.5 Flash in production and Antigravity 2.0, a multi-system agent orchestrator. Microsoft is making a parallel push using its cloud and balance sheet. The segment Anthropic owns most clearly right now — AI for builders — is suddenly the segment three labs are competing for the hardest.

Why it matters for you: If you've been paying for Claude Pro or Max because it's the only thing that's worked for serious building, this is the month to actually benchmark alternatives. Not to switch. To know what your fallback is when you need one.

Sources: CNBC on the coding AI race · Build Fast with AI — June 5 recap

🌶️Hot Take: Agents are the new compound interest

Most founders ask the wrong question about velocity. They ask "how do I move faster" and reach for the usual answers. Better tools. More focus. Smarter shortcuts. None of them work the way they used to.

Here's what's actually happening now. When you build with an AI agent in the loop, the work compounds in a way that nothing in software has compounded before. The agent that brainstormed Spire with me on Thursday wasn't smarter than I am. It was prepared the way a teammate I'd worked with for years would be. Every prior session, every preference, every "no, I don't do it that way" was already in its context. It knew the shape of my business before I asked.

The 4-hour ship wasn't a stunt. It was the receipt for six months of teaching a stack of agents how I think.

This is the part I want you to sit with. The founders winning the next decade aren't going to be the ones who write the most code, or raise the most money, or have the best ideas. They'll be the ones who invested early in agents that work like teammates. Every hour you spend now teaching an agent your taste, your constraints, your workflow earns interest. Compounded daily. Paid out the next time you sit down to ship.

I'm not telling you to fire anyone. I'm telling you to start hiring. The unpaid, always-on, gets-better-every-week kind. It's the most absurd leverage I've ever had access to.

The cheat code isn't speed. It's the agents you built before you needed them. Six months of small, daily, boring discipline turned into a 4-hour ship on a random Thursday. That isn't magic. That's compounding showing up on time. The same kind of compounding that's available to anyone willing to do the quiet work, every week, when nobody's watching.

Let me tell you what actually happened, because the "4 hours" headline misses the real story.

Thursday, 6:00 PM. I open a Claude session and say something like: "Help me brainstorm a small iOS game I could ship this weekend. Constraints: no online multiplayer, no servers, has to be fun, has to fit my brand." Twenty minutes of back-and-forth, three ideas surface. One sticks: a one-tap stacking game. Tap to drop each block, line it up perfect, watch the tower climb. Miss and the edges slice away. The tagline writes itself: How high can you build?

6:30 PM. New project. Claude Code spins up the Expo skeleton. I'm not writing the boilerplate. I'm picking the gameplay feel, naming things, deciding the failure state. The agent is doing the typing.

8:00 PM. Playable on my phone. The first three levels work. Tower falls when you stack too hot, comes back together when you place a block clean. Combos when you stack perfect-on-perfect. I'm laughing at it.

9:30 PM. App icon, screenshots, App Store Connect metadata, RevenueCat configured for the IAPs (4 coin packs + an Unlock Everything entitlement), privacy policy hosted on sunventures.studio. Every one of those used to be a half-day of paperwork. Each was 15–20 minutes with an agent walking me through it.

10:00 PM. Binary uploaded. Submitted to review.

That's the whole story. 4 hours.

But the actual story is six months long. Every Claude Code session before this one taught my agents how I name files, where I put my privacy policies, what bundle ID conventions I use, what my tone is in App Store descriptions, what fonts I default to, what I let RevenueCat decide and what I override. Thursday was just the day I asked them to use all of it at once.

Spire status: under review at the App Store as of this writing. Could go live this week. Will probably go live looking nothing like a "4-hour build" because the agent stack underneath it is six months deep.

Also this week: I shipped dedicated privacy policies for every app at sunventures.studio (Caption Coach, Tabflow, ToneSync, MothersTouch, Spire). I hardened the Cloudflare Worker proxy fronting the Caption Coach API: rate limits, request caps, model lock. I rebuilt the Sun Ventures homepage so Spire shows up as product #6 in a dual-phone hero. None of it would have taken what it used to.

I didn't ship a feature this week. I shipped a whole new product. With agents that have been ready for it for months.

⏱️The Hidden Hours: What this week actually took

The real ledger, week 2 of this section. Use it as a benchmark, not a target. The point is the foundation, not the math.

🚨 Fires · ~3h
Cloudflare 2FA, Caption Coach proxy hardening, broken legal link cleanup

🤖 Active products · ~8h
Spire build + submission, Caption Coach proxy work, website Spire integration

🎥 Content (@buildwithkp) · ~12h
Agent-assisted scripts + edits for IG and TikTok, plus the Day 1 and Day 3 videos

📩 Community + outreach · ~4h
DMs, replies, prospect outbox

✍️ Newsletter + writing · ~3h
This issue

💛 Off-keyboard life · ~25h
Family, gym, sleep, walks, decompression

That's roughly 30 productive hours and 25 off-keyboard. The remaining 113 hours are sleep and the small admin in between. The math closes.

This week the foundation was different from last week's. Last week was infrastructure I controlled. This week was agents I'd trained. Both are the same kind of work. Boring, invisible, paid forward.

Half an hour. Pick a workflow you do every week that you don't love doing. Doesn't have to be glamorous. Writing app descriptions. Drafting outreach emails. Summarizing meetings. Naming files. Choosing thumbnails. The boring repeatable thing.

Step 1. Open Claude (or your favorite AI). Tell it the workflow. Describe exactly what good output looks like. Show it two or three examples of your past work in that area.

Step 2. Tell it the constraints. Your tone. Your length. Your "do nots." Specific words you never use. Things you always check before publishing. The texture of your taste.

Step 3. Save it as a Claude Project (or a system prompt you reuse). The point is durable memory. The agent that helps you next week should remember everything from this week's session.

Step 4. Use it once. This week. On a real task. Notice where it nails the texture and where it doesn't. Add the correction to the system prompt for next time.

That's the whole loop. One agent. One workflow. Half an hour of setup. Then it gets a little better every time you use it.

Not because it'll feel like a breakthrough this week. Because the work compounds. Six months from now, the agent you set up this Saturday will be running parts of your business for you while you sleep. The only thing standing between you and that outcome is showing up. Once a week. Consistently.

If you do this and the agent surprises you, hit reply and tell me what it figured out. I'm collecting these.

That's Issue #4.

Issues #1–3 ran a trilogy: ship right, constrain right, build right. Issue #4 starts something new. How a solo founder compounds. Not by working harder. By teaching agents to work the way they think.

I've been planning something for a while. In the next week or two, I'm going to start a long-form YouTube channel. Same brand, same voice, much more depth. About exactly the thing I just wrote about: using AI agents across every part of running a tiny business. Content planning, lead gen, product building, marketing, teaching. The whole stack, in the open.

Here's the honest version. If you've been waiting for permission to start the thing you've been thinking about — to launch the side project, ship the first version, write the first email, post the first video — this is it. The work isn't glamorous. It's small choices, made consistently, when nobody's watching. But consistency plus the right agent stack compounds into things that look impossible from the outside. Six products from a kitchen table. A game in four hours. A newsletter that found you. None of those were lucky. They were boring, weekly, disciplined effort that one Thursday started paying out.

I'm going to film exactly what that work looks like, week over week, so you can see for yourself. Nothing flashy. Just the real thing.

If Issue #4 landed for you, forward it to one founder who's still doing everything themselves. And hit reply with the workflow you're going to teach an agent first. I want to see them.

See you next Sunday.

Quietly, deliberately. One Saturday at a time.

— Kalpesh Mistry
Founder, Sun Ventures
sunventures.studio