Founder intuition, A/B testing, and building Rewind AI

Executive overview

Founders accumulate a dense, hard-to-articulate signal from across their company. When that gut says one thing and an advisor says another, ignoring it is the most common source of regret.

Dan Siroker traces this lesson from the Obama campaign through Optimizely's rise and over-hiring to the founding of Rewind AI — a passive memory tool built on the insight that technology can transcend biological limits.

Trust your conviction: the times you override your intuition and fail are the ones you remember longest.

The Obama campaign and the power of A/B testing

• Obama was third in the polls; campaign manager David Plouffe embraced a "risk-reward" mindset — doing what every other campaign did would produce the same third-place result.
• A single A/B test on barackobama.com — swapping a presidential photo for a casual family photo — drove an incremental $40 million in donations.
• The lesson: A/B testing is just the scientific method applied to real decisions; the insight isn't the test, it's lowering the friction to run one.

Building Optimizely: democratising experimentation

• Optimizely was built around the product Siroker wished he'd had on the campaign — easy, no-developer A/B testing via a JavaScript snippet and a visual editor.
• Competitors like Adobe required large professional-service contracts and multivariate tests; Optimizely flipped this by supporting simple sequential A/B tests first.
• Series A ($28M via Benchmark) → Series B ($57M) followed rapid early growth; ARR reached ~$7M within three years.

The over-hiring trap

• After raising large rounds, Siroker didn't enforce headcount discipline: new executives hired whole teams without challenge.
• Hundreds of employees meant most of his time went to alignment and management, not product — the work that energised him.
• He fell out of love with the business; disengaged founders struggle to attract and retain great people, compounding the damage.
• Lesson: raising more than you need is defensible; spending more than you need is not.

Trusting founder intuition

• The pattern he regrets most: gut says A, a board member or executive says B, the team does B, B fails.
• Founders synthesise engineering, customer, sales, and market signals into intuition that is real but hard to articulate.
• At Rewind he now defaults to his conviction; most wrong calls are reversible two-way doors.

From hearing loss to Rewind AI

• Wearing a hearing aid for the first time revealed how much had been lost — and sparked a decade-long question: what other biological limits can technology transcend?
• First product (Scribe) transcribed meetings only; users immediately asked for broader capture, and Apple Silicon made comprehensive on-device recording viable.
• The pivot to Rewind — capturing everything passively — created a new category, not a better note-taking app.

Building new categories

• Most note-taking tools require foresight, search effort, and active input; Rewind is passive, comprehensive, and imperceptible.
• Signs a new category is needed: people cobbling together mismatched tools with "duct tape and bailing wire" to solve a job to be done.
• Start with the problem, not the technology; falling in love with GPT-4 and then hunting for a problem inverts the correct order.

Raising the Series A publicly

• Published an unedited seven-minute pitch on social media; 2 million views, thousands of investor offers, hundreds of qualified meetings.
• Motivation: build user trust around a privacy-sensitive product, and handle inbound investors at scale without individual meetings.
• Extended runway from three years to six; primary growth driver remains word of mouth at 15% MoM ARR growth.
• Transparency as a competitive advantage: a startup can do things a Fortune 500 CEO cannot, and that asymmetry builds differentiation.

Advice for founders choosing a problem

• Painkiller vs. vitamin: people only change entrenched habits for problems that are genuinely painful.
• Ask honestly whether you would change your behaviour to use the product you're building.
• Choose a problem you'd pursue even if it failed — because if it succeeds, you'll be living with it for 10–20 years.
• Flywheels and competitive moats matter eventually, but only after you've built something people actually want.