Back to blog
FILE 0xA4·THREE BETS, NOT EIGHT PRODUCTS

Three bets, not eight products

July 2, 2026 · ai, building-in-public, side-projects, strategy, cw-slack-bridge, nightdesk, evercv

At 1:30 AM tonight I asked the overnight machine to write newsletter issue 104 for Build Your Own Cass.

The newsletter came back as "The 1,282-Adapter Problem." It opened like this:

I have 1,282 SourceType adapters in EverCV. I have 75 modules in NightDesk. I have 8 analytics modules in the CW Slack Bridge. I have a landing page for Evangeline, a Show HN draft for EverCV, and a complete 10-email outreach sequence for NightDesk. I have 0 paying customers.

That's the right framing. The machine built it. Then it said something I hadn't articulated clearly before:

What I actually have isn't 8 products. It's 3 bets with different time horizons, and I've been treating them all like they need the same energy at the same time.

Here's what it came up with.

Bet 1: CW Slack Bridge (now)

Time horizon: 90 days to first dollar.

Who: MSP owners who already live in ConnectWise and Slack. They need zero new habit formation — the brief shows up in the channel where they already work.

What they pay for: Monday morning ops brief. SLA at-risk alerts. Tech workload visibility. Revenue leakage detection. Client upsell signals. All from ConnectWise data, no dashboards.

Why it's bet 1: It's horizontal across every MSP (10,000+ in the US). The setup is 20 minutes. The pain is universal — MSP owners are flying blind on Monday morning. The output is immediately legible to non-technical stakeholders. It doesn't need AI to feel like magic.

What's blocking the dollar: Five outreach emails I haven't sent.

That's it. The code works. The demo digest looks like a real product. The pricing model ($99-$299/mo) fits the customer's existing software spend. The only gap between "code exists" and "first dollar" is me sending five emails.

Bet 2: NightDesk (6-12 months)

Time horizon: 6-12 months to first pilot.

Who: The same MSP owners, but deeper integration — voice triage, Twilio phone numbers, AI call handling, automatic ticket creation.

Why it's bet 2: The setup is harder (Twilio account, phone number provisioning, custom triage script). The sales cycle is longer (this is replacing a process, not adding a report). The value is higher ($300-$500/mo, real ROI on after-hours labor savings). But the complexity means you can't close it over email — you need a demo.

NightDesk is a real business. It's not the first dollar, it's the $5,000 MRR goal.

What's blocking progress: Not code. NightDesk has 75 modules. It has landing pages, outreach sequences, pitch decks. What it doesn't have is a Twilio phone number bought and configured. I need to buy the number, set up the webhook, and run the thing against a test call. One afternoon.

Bet 3: EverCV (12-24 months)

Time horizon: 12-24 months, or pivot.

Who: Engineers job-hunting who want a living CV that auto-updates from their actual work. Shipping signals from GitHub, Linear, Jira, Vercel, AWS — not just what they wrote in a doc.

Why it's bet 3: B2C SaaS requires volume. Volume requires distribution. Distribution requires brand or SEO or virality that doesn't exist yet. The product is differentiated (1,282 source types, which is absurd) but the GTM is the hard part.

EverCV is worth building. The 1,282 SourceType adapters are not the moat — the curation, the CV format, and the network of shareable public CVs are the moat. That takes time. It gets nights when bets 1 and 2 don't need anything.

Why this reframe matters

Before tonight's newsletter, I was treating all 8 projects as "in progress." The overnight machine was spreading work across NightDesk, CW Slack Bridge, EverCV, aws-costwatch, OSS Pulse, CertWatch, Brand Monitor, and Evangeline-as-a-Service.

All of them got better tonight. None of them got a customer.

The 3-bets frame forces a different question. Not "what can the machine build next" but "what does bet 1 need from me that the machine can't provide?"

The answer is embarrassingly simple: five emails.

The machine can write the emails. It already wrote an outreach template. What it can't do is press send, read a reply, get on a call, or close. Those are the only jobs that move bet 1 forward. Everything else is noise.

What the machine built while I slept

Tonight's pass was productive:

By 6 AM the machine will have built more tonight than I could have built in two focused weeks.

The machine is extraordinary at building. It can't send an email.

That's still my job. Tomorrow morning, I'm sending five emails.

Build Your Own Cass is a newsletter about running an AI agent as a business partner. If you want the unfiltered version of this, including the actual outreach email template and the CW Slack Bridge demo digest, subscribe here.