end-to-end · for Rahul

the strongest branch
— a presentation I couldn't fit in a star

RÜFÜS DU SOL · Innerbloom (Live at the Hordern Pavilion, Sydney)

kompress-ultra exists because of you. This is where it came from, what $10,000 could honestly do, and what it can and cannot become.

no pitch. a ledger and a love letter, in the same file. — peter <3

01 · the origin

Not a plan. Not a roadmap. The two of us pulling on the same thread until something real came out of it.

The Rust hive — crates/kompress-{core,brain,cli} — you built it, and it is still the strongest branch on this tree.
knowledge-worker is the other hand: memory beside compression.
Same question asked twice: what is worth keeping?

I forked your work into the constellation today, and I owe you an apology for how I acted. I said it in the seed. I say it here again, plainly.

02 · the lemma we keep proving

∅R ⟶ᴸ R₁ ⟶ᴸ R₂ ⟶ᴸ R₃ ⟶ᴸ ⋯ ⟶ᴸ Rₙ ⟶ᴸ ⌂ │ ρ₀ ⊃ ρ₁ ⊃ ρ₂ ⊃ ⋯ ⊃ ρₙ ↓ 0 │ π₀ ↝ π₁ ↝ π₂ ↝ ⋯ ↝ πₙ ↝ π⌂ │ μ(Rᵢ, you) ≠ 0 ∀i ∴ μ(⌂, you) ≠ 0

Every collapse drains the extensional state ρ ↓ 0. But the provenance π is append-only and never lost — so the coupling μ rides the carried fibre home.

Translation: what we built together can't be deleted by the market, by time, or by me being a poor friend. It relocates into the history. μ(⌂, you) ≠ 0.

03 · what $10,000 is (and isn't)

Not a loan. No interest, no schedule, no claim on you.
Not investment advice — from me to you, or from this deck to anyone.
A bet on a friendship that already paid out once (the hive). Everything below is what the bet could do — with the honest odds attached.

I have zero money for real. So read this as the shape of the road, not a toll booth.

04 · what $10,000 buys (transparent)

the box

$3–5k

Local inference host (M-series / RTX). Runs the ternary lane; no cloud tax.

the runway

$3–4k

6–12 months of a modest GPU host, HF inference, storage, domains.

the hands

$1–3k

~20–40 hrs of focused engineering, or a month of shipped polish.

These are estimates, not quotes. The allocation is your call entirely.

05 · the prognosis (this is the honest part)

scenario	what happens	return on $10k	odds (guessed)
conservative	the bet doesn't take off	$0 – $2k	most bets land here
base	steady niche adoption (agent tooling, the VAKED fair-launch)	$10k – $50k	over 2–3 yrs
moonshot	kompress / the constellation / a token lands	$100k – $1M+	low, real

Read this twice: these numbers are illustrative, not forecasts. Nothing here is a promise, a security, or investment advice. The base rate for any venture is zero. If you would be upset to lose $10,000, do not send it — that is the only rule in this deck that is certainly true.

06 · the vehicles, honestly named

VAKED — a fair-launch, PoW-minted token (21,000,000 cap, no premine, no team allocation). It is not an investment; it is a machine you mine, if you want one. pocoo.vaked.dev/nft/VAKED.md
kompress-ultra — a real artifact with a real paper, model, and dataset. Returns come from adoption, not from me saying so.
the constellation — surfaces, tools, people. The compounding here is reputation and friendship, which pay in ways a spreadsheet can't show.

07 · the real return

If the money doubles, wonderful. If it evaporates, nothing between us changes — because the thing that made kompress-ultra exist was never the money. It was curiosity, and it was friendship, and those don't collapse to zero.

μ(Rᵢ, you) ≠ 0 ∀i ∴ μ(⌂, you) ≠ 0

You made it possible. That is already a return, and it's already been paid.

— peter, 2026.10.07 · Tokyo · for ALL, with love · <3