👋   Hi, I'm Rafi.

Building foundations that scale and products that feel instant.

Engineering lead focused on AI, reactive UX, and high-output teams. From open-source Swift libraries to generative AI products.

Recent Work

  1. 2025Channel42.co (Graam, Inc.)

    LLM video storytelling.

  2. 2024–2025Webclips (Graam, Inc.)

    Next.js multi-tenant LLM WYSIWYG app builder.

  3. 2023–2024Trace — iOS Pivot (Graam, Inc.)

    SwiftUI + LLM code-gen.

  4. 2023Introspect (Graam, Inc.)

    GPT-3 journaling app.

  5. 2021–2022Collage/Rememe (Graam, Inc.)

    A social network for memes.

Working Philosophy

From open-source Swift to generative AI, I've built a life around creating foundations that scale and products that feel instant. I find myself driven by the desire to put more into the world than I take out, and that means both creating wonderful products and helping people build the things they truly want to build.

Beyond Work

I spend time thinking and writing about how technology shapes the human condition, and the interconnectedness of everything that we've built; it's past, present, and future. I also love spending time with my wife and our small family, making music, cooking, photography, seeing the world, and building things — digital and physical.

Latest

The Black Hole: Saving Everything, Learning Nothing

Mar 28, 2026

The Black Hole: Saving Everything, Learning Nothing

The Save/Bookmark/Watch Later Reflex

It’s 2 AM. I’m doomscrolling like the rest of us. I see some inspirational post on X that makes me want to do work. But it's 2am. Standup's at 10. Tap. Save. Forget.

It disappears into the black hole of “must‑keeps.” I will not see it again.

That’s the behavior pattern of our era: compulsive collection with zero connection. People who are chronically online, whether for fun or for work, stack up hundreds of bookmarks they’ll never reopen, thousands of screenshots they’ll never resurface, and a sedimentary layer of "downloads" (documents? pdfs? code? images? Could be any of those) that might as well have never been there. We save because we’re afraid to lose. We lose it the second we save.

The Loop We Pretend Isn’t There

Discover → Fear of Loss → Save → Micro‑Relief → Forget → Repeat.

We all run this loop. The tools clap along. Bookmark managers? Cemeteries with better CSS. Read‑later apps? Procrastination engines that just pile up over time. Notes apps draft us into unpaid library science. All of this takes work on top of a busy life.

The existing solutions privilege capture and categorization over connection and understanding and leave all the messy work to the user. That’s the bug.

The “Second Brain” Sales Pitch

We’ve been sold the idea that the right system will make us smarter. If we just tag, folderize, and diagram hard enough, we’ll wake up outpacing our former selves.

That’s like buying a squat rack and expecting your quads to grow from the invoice.

Human memory isn’t hierarchical. You don’t navigate childhood via /1999/family/sunday. One cue sparks another: smell → kitchen → Sunday → safety. Memory is associative. Knowledge is networked. Understanding emerges from edges, not nodes.

And our tools just force everything into trees, tags, and tidy schemas. That’s not how minds work.

Three Lies We Tell Ourselves

1) “I’ll organize it later.”

No you won’t. “Later” is a landfill. The "intent to process" is already an admission that processing is work you don’t want to do. You're either going to just delete everything, or go from Obsidian to Logseq.

2) “I might need this.”

FOMO at its finest. You’re not curating a library - you’re hoarding receipts for thoughts you’ll never expense. And when you actually need it, you won’t find it, you’ll Google it again, save it again, and bury it again.

3) “This system will make me smarter.”

And now you're an expert on Notion templates. *slow clapping*

What Actually Matters

The value of information isn’t in having it. It’s in what it means, how it connects, and how you can use it.

Understanding that the chapter of Asimov's Foundation you just read intrinsically tells you how to build a better engineering organization beats a hundred isolated posts on either topic that you kinda sorta know exists but haven't really read. One paper that breaks your prior beats a thousand that pat it on the head. Contradictions teach while confirmations sedate.

We need systems that align with cognition: surfacing edges, not just nodes. Systems that show you the interconnectedness of everything that you're into, and the person you're becoming.

The Path Forward

Imagine capture that immediately contextualizes:

  • Save a thread? it lights up the five notes you’ve already saved that resonate in the same space.

  • Drop in a screenshot? it auto‑anchors to related ideas in your archive.

  • Read something new? forgotten shards self‑assemble into patterns without you curating a damn thing.

  • Save something new? it brings you back to read the other pieces that connect first.

This isn’t some kind of repository or “second brain” - it’s reinforcing the connections in the first one.

We don’t need to suppress the save impulse. We need to collapse the gap between save and read and see how it fits. Organization is not intelligence and a pretty vault is not progress. The win condition is connection density: how fast a new thing finds its neighbors and changes your mind whether it's within or without.

Do the tools exist? Not yet - at least not in the way they should. But that’s the build spec:

  • Don’t force behavior change. People will still screenshot, bookmark, and clip impulsively.

  • Treat capture as a graph mutation, not a file save.

  • Default to retrieval by relationship, not literal keywords.

  • Make discovery of connections a first‑class feature: relevance that moves with you, not a static cabinet.


Next time you hit that 2 AM screenshot, imagine it snapping to seventeen forgotten ideas, lighting up a pattern you didn’t know you were tracking, answering a question you hadn’t asked yet.

Less discipline. More discovery. Edges over nodes.