DealForge autonomously sources, scores, and writes investment memos on venture deals. Stop manually hunting.
1,180+ deals tracked · 22 AI investment memos · Updated daily
I built a desktop AI that picks between 10 founder voices (Collison, Benioff, Lütke, Chesky, Altman…) for sales questions. Here's what I learned about "founder voice" as a product.
I kept asking Claude for sales advice and kept getting generic "here are 5 tips" output. That's because Claude defaults to a committee average of everything it's read — LinkedIn, Medium, bad blog posts, and the actual good stuff all mashed into one voice. So I built a desktop app that makes you pick which founder you want to hear from. Ten voices loaded as separate skill files: Collison (developer-first, API-led), Benioff (enterprise trust), Lütke (underdog framing), Chesky (identity-driven), Huang (platform bets), Altman (freemium-to-enterprise), Amodei (substance over pitch), Levie (competitive positioning), Butterfield (product-led growth), Lemkin (operator playbook). You type a situation. The app picks the right founder. You get the answer in their voice — their priorities, their framing, their style, based on their public writing and talks. **Three things I learned building it that I didn't expect:** **1. The "voice" isn't the style — it's the question it forces you to re-ask.** When I asked Collison-mode "how do I convert more trial users?" the first thing it did was reject the framing: "you're optimizing the wrong end of the funnel — the trial is fine, the activation step inside the trial is broken. What does the first 10 minutes look like?" That reframe came out of Collison specifically because his public writing obsesses over first-principles re-framing. Benioff-mode would've given me an enterprise sales playbook instead. The value of a voice isn't the output style — it's the question the voice forces you to ask before answering. **2. Ten voices is way too many for some questions and exactly right for others.** Cold outreach? You want Lütke. Pricing a big contract? Benioff. Naming a feature? Chesky. Deciding whether to raise? Altman. I added a router that picks for you so you don't have to know. The router turned out to be the real product — most users don't actually know who they want to hear from, they know what problem they have. **3. Running it lo
The startup identifies a high-value niche in AI-driven mentorship by prioritizing the 'philosophical reframing' of business problems over generic output. Despite the clever product insight and 'router' logic, the project faces extreme platform risk and lacks a clear defensive moat or verified traction, making it a speculative seed-stage play.