Nature

Update to the DYBL primitives library. Two new libraries, Breath and SeasonArc, now sit with EternalSeed and BreathEngine (internal names). Breath: what a pool with no end date can afford to pay. SeasonArc: a rising arc to the last payout. Pre-testnet, not audited. Known limits are in each KNOWN_ISSUES.md. github.com/DYBL777/dybl-prim… 🧵1/3

We call it the Autonomous Distribution Rate Problem. It exists in every protocol that holds funds and owes obligations. To our knowledge no production protocol answers it autonomously and forward-projecting. The honest limit, documented in the library itself: it can return zero, and zero is ambiguous. Structural insolvency and a solvent pool sitting exactly at its floor both return zero. So the insolvency predicate ships alongside it. Pure Solidity library. Five functions. Stateless. No storage, no dependencies, no external calls in the math. The reference host runs on @Chainlink CRE-style orchestration: a periodic workflow advances the host, the host derives a rate. Part of the DYBL primitives library (alongside Compounding Reserve). github.com/DYBL777/dybl-prim…

Yesterday's post was about protecting a pot. This one is about the question that comes next. Once a pool has a protected floor, something has to decide the rate. Distribute too fast and you breach the floor before maturity. Too slow and you are hoarding capital that should be working. Most protocols typically answer this with a fixed rate set at launch, or a governance vote every few weeks. Neither is a control loop. One ignores what actually happened, the other is too slow to respond to it. Solvency Autopilot (internally BreathEngine) answers it on demand, every period. Give it four numbers about your pool (current stock, obligation floor, periods remaining, estimated inflow per period), plus your rollover fraction and your own rate ceiling. It returns the maximum rate you can distribute this period without breaching that floor at maturity, on its own projection. A 24-iteration binary search where every candidate is verified against a forward projection of the full remaining horizon before it is accepted. This came out of the seed work. The protected amount was the first half of the problem. The rate is the half most protocols leave unsolved. 🧵 1 of 2

Most pooled protocols have the same quiet flaw. They distribute until the pot is gone. No floor. No protection. Just distribution until there's nothing left. Compounding Reserve (the mechanism internally known as Eternal Seed) solves this. It locks a protected amount inside the pot for the duration of the season. That amount cannot be distributed. It sits there, season-locked, compounding draw on draw. Two modes: a floor the pot can never fall below, or a recurring flow that rolls a fraction of each prize pool back in. Either way the protected amount is untouchable by governance during the season. Defined exit paths exist at settlement and dormancy. Permanent within the season. Not beyond it. Pure Solidity library. Stateless. No external calls in the math. Extracted from a fully tested suite and behaviour-matched against the source contract. Designed to compose with Chainlink Data Feeds and CRE-style orchestration. The host supplies state and policy, the library supplies the capital-protection math. Part of the DYBL primitives library (alongside Solvency Autopilot). github.com/DYBL777/dybl-prim…

The DYBL primitives library is live. Two Solidity libraries, pure functions, no state, no external calls. Eternal Seed for capital retention, BreathEngine for solvency-aware distribution. Both extracted from a working prediction market and proven byte-equal against it. github.com/DYBL777/dybl-prim…

Prediction markets are having their moment. DYBL is building the capital-retention and solvency primitives underneath them. Eternal Seed and BreathEngine, so pools stay solvent and compound instead of resetting every season. Demonstrated inside a Chainlink Data Feeds + CRE settlement game. Primitives library goes up soon.

Spent the morning on documentation. Three files in BullsEthCRE had drifted out of date. Fixed. 432 tests across 24 suites, nine invariants running in CI, three Highs still open and listed in full. The front page now says plainly that the repo is frozen, and why. It's a reference implementation used to verify extractions, not something heading for deployment (by me). None of this is new work on the primitives. The eternal seed and the breath engine have been built alongside the game for months. Their library goes up next. github.com/DYBL777/BullsEthC…

A correction to this. The numbers hold. 70.8KB, both flagship features cut, still 23KB over. What I got wrong is the conclusion I hung on them. "The split was never optional" reads like the contract can't be deployed. There is a deployable path: deployed libraries and satellite contracts. It's a long refactor with every test to redo, and I haven't built it to prove it. A hill, not a wall. So freezing it is a choice, not a verdict. One person, 9 months in, and I'd rather spend the next stretch on the primitives than on a rebuild that ends with a demo. The maths only ever said that trimming the two biggest features isn't enough. Test count is 432 now, not 425. The suite grew.

Been building the primitives and their Foundry test suites alongside the game for around 3 months. The breathing engine is extracted and proven to behave identically to the version inside the game and nearly ready for its own repo, then making it adaptable, so it works for a plain pool with no OG endgame at the end.

Sizing experiment on the monolith. 70.8KB deployed, 46KB over the EIP-170 limit. So I cut dormancy entirely and rebuilt: 11.7KB back. Also the whole OG loyalty system: 11.7KB. Both flagship features gone and it's still 23KB over. No amount of trimming reaches deployable. The split was never optional. The math just made it official. So the game freezes as a fidelity anchor: 425 tests, 9 invariants, fuzzed across resets and dormancy. Every primitive extracted from here gets verified against it, byte for byte. Solvency-aware distribution, the eternal seed, a dormancy waterfall. And three answers to cold start: milestone-gated investor capital, skin-in-the-game staking, pregame commitment. The primitives were always the product... Now comes the simplicity

Dormancy and emergency reset now under fuzz. 432 tests. Twelve actions, nine invariants, 12,800 calls per property. Two bugs in my own handler before the campaign ran clean. Reset and both dormancy calls are owner only and wired the handler as the forwarder. Both destructive paths silently dead, three call sites, one root cause. Then dormancy sat behind a dice roll while being the only legal exit from a stranded season. The suite's own progress assert caught that one on its first outing: a stranded run failed CI instead of passing quietly. github.com/DYBL777/BullsEthC…

Invariant fuzzing, first pass. Up to 430 tests. Seven properties over a live season handler. Balance covers every allocated pool, seed release stays inside the ratio ceiling the solvency proof assumes, treasury never exceeds holdings. First campaign came back green on all seven. It had also completed zero draws in 12,800 calls. Warp action moved the clock blindly, and a few of those before the first buy shut the pick window. Unrecoverable from there, so every property passed trivially, forever. Fixed by making time phase aware, and by counting why each action did nothing. Silent returns are right, but they made "not applicable now" and "can never work" identical. Now nine draws in one sequence, cutoffs accepted nine and failed zero. The campaign asserts its own progress, so a stuck run fails CI rather than passing quietly. Live season only. Second pass adds reset and dormancy, where the ordering bugs live. github.com/DYBL777/BullsEthC…

Two more test files. 415. These cover the money paths that only exist after an emergency reset. If a draw gets voided, players who paid in but never got to play can claim it back, and whatever nobody claims gets cleared out after thirty days. While the game is still running unclaimed money goes back into the prize pot, so it gets paid out to players later. Once the game has wound down there's no season left to fund, so it goes to the protocol. Same function, opposite recipient, decided entirely by what state the game is in. github.com/DYBL777/BullsEthC…

Two more test files... total, 404. The first covers the draw-30 bonus fund. Three percent of every draw's prize money is held back so the final draw is the biggest of the season. The second is the solvency check. The contract refuses to run a draw if it doesn't hold enough to cover what it owes. One test forces that check to fire. One proves it stays quiet when everything's fine. Prove it works, then prove nothing breaks it. Only the pair means anything. Next up: the two refund paths that only exist after an emergency reset, and the sweeps that close out a finished season. Both move money the accounting has to account for. github.com/DYBL777/BullsEthC…

Distribution tests done. 390. The suite knew how many winners each tier had. It knew the money added up but had no idea what any winner actually got paid. Nothing checked that T1 pays more than T2 pays more than T3. Swap the tier numbers round and every test still passes and counts fine. Money conserved. Draw completes. Only the payouts are wrong, and nothing was watching. Fixed. Plus dust, batched payouts past the 200 cap, and ties landing exactly on a cutoff. github.com/DYBL777/BullsEthC…

Three rounds since the last update. 379 tests up to now. Weekly OG lifecycle first. Buy every draw for a season, one miss ends it. The qualification threshold equals the season length, so it is only reachable on the final draw. Then ten revert expectations that had no error type on them. Each was passing on any revert rather than the intended one. All ten turned out to be passing for the right reason anyway, but nothing proved it until now. Then a branch sweep found three places that take a qualification back and tests for none. Two are structurally unreachable. The third needs the operator to wind the game down on the exact draw someone qualifies. That one has a test now. github.com/DYBL777/BullsEthC…

Treasury and automation tests done. 360 total. Treasury withdrawal is the only path where owner could move money to itself. Four locks guard it and none had ever been checked. The automation side is whatever pokes the contract so the game keeps moving. Chainlink Automation or CRE, same entry points, same failure mode: if it reports wrong, nothing errors. The season just quietly stops. Three review rounds. Every finding was in my tests. None in the contract. Weekly OG lifecycle next. github.com/DYBL777/BullsEthC…

Governance timelocks covered. 41 new tests, suite now at 304. A timelock is a forced wait between announcing a change and being allowed to make it. Announce, wait seven days, then act. Anyone watching gets warning. Two review rounds found four problems. All four were the tests. None in the contract. Best one: checked that a bad feed address gets rejected. It does, but with a blank error, not the named one I expected. The named check I thought I was testing never runs at all. github.com/DYBL777/BullsEthC…

Breathing engine testing complete. Took longer than everything else combined. It's the hardest thing in the suite. It sets the payout rate every draw so the pot stays solvent to the end of the season. Every other mechanism does one thing. This one has to keep doing its job while the ground moves. Governance timelocks next.. after I touch sand and sea ;) github.com/DYBL777/BullsEthC…