#Sami Abuzarifa

Product strategy

Discovery, roadmaps, and the judgement to kill the right things early. I have done this across FinTech, blockchain, and productivity, usually as the first product hire or the person who has to make zero-to-one actually happen.

Selling what I build

Scoping with enterprise, SME, and government buyers, writing the proposals, running the negotiations, then owning delivery on whatever I just promised. Across XHAI Studios and LuxTag that came to more than $50 million in revenue.

AI and LLM products

Hands-on with OpenAI, Claude, and Gemini. Retrieval, evals, and agent orchestration shipped to real users, including the unglamorous part: our dense embeddings could not find our own product codes, and hybrid retrieval took recall on the internal set from 61% to 89%.

Positioning and launch

Pricing, messaging, and getting a team to say out loud who the product is for. I judge a launch by the pipeline it creates six weeks later, which is a slower and more honest number than the one everyone celebrates on the day.

Leading the team

Cross-functional teams and dual-track delivery, with enough process to keep a room aligned and little enough to keep it moving. I have hired these teams, run them, and had to shrink one, which taught me considerably more than the hiring did.

Termind \u2197

Joins web analytics, product events, and Stripe revenue, then says what changed and why. For founders who have every chart and still no answer.

GistAI \u2197

Most chat widgets wait in the corner to be clicked. This one reads dwell time and scroll depth, decides for itself when to speak, and answers only from your own content, with citations.

Bookbind \u2197

One sentence in, a sellable book out. Writes the pages, draws the art, and packs print-ready files that pass the checks at Amazon KDP, IngramSpark, and Lulu.

Flint \u2197

You talk, it writes. Native iOS and Android voice notes with no recording limit, paid for once instead of rented. Audio stays on the phone.

Komiks \u2197

Make a comic, not just an image. Turns a story pitch into a finished issue with a consistent cast, lettered panels, and a cover.

Shipped \u2197

Founder Circles. Puts under-followed builders on X into small private rooms with real peers, without the engagement pod mechanics.

Discovery is the bottleneck now.

AI took build time from weeks down to afternoons and left discovery exactly where it was. I have watched a team ship four versions of the wrong thing in the time it used to take to ship one. Speed only helps you if learning keeps pace with it.

Forty labelled examples beat a month of user calls.

That is roughly what it took before I understood where one of our AI products was actually failing. Evals get filed under QA and run like a dashboard. Run them like discovery and they start telling you what people are really asking for, often before those people can put it into words.

Ask what actually changed for the user.

Building got cheap enough that shipping proves very little on its own. The skill worth having now is causal: name what moved, name who it moved for, and show your working. Most roadmap arguments I have sat through would have ended in ten minutes if one person in the room could do that.

Your tool description is the new homepage.

In AI distribution a model reads a sentence or two about your product and decides in that moment whether to surface you at all. Name the trigger, the user, and the output, in that order. I have rewritten more of these this year than landing pages.

Keep one original thought.

Every team now runs the same frameworks inside the same tools, prompted much the same way, and arrives at the same answer within a week of each other. Use the framework by all means. Then argue with it, because the argument is the only part a competitor cannot get by buying the same subscription.