Commit to the Weaker Assumption, Then Optimize to the Limit
Once you've paid the price of a compromise, stop hedging and exploit it fully.
- Difficulty
- Moderate
- Time to result
- ~months to results
- Steps
- 6
- Confidence
- 80%
Proof of stake gives up the objective, from-scratch verification that proof of work provides: a token has no thermodynamic weight, so a newcomer must first ask the world whether they are on the right network. That is weak subjectivity, and Yakovenko's argument is that once you accept it you should stop pretending otherwise. Trust-on-first-use is already the standard security model for the internet — a browser trusts a root certificate, establishes a session, and continues. Having conceded the same thing, every remaining constraint inherited from proof of work is voluntary. Removing them opens the design space to a single question: how fast can this communications network go? The method generalizes — identify the guarantee your design already forfeits, verify the cost is sunk, then delete the constraints that only made sense under the old guarantee.
Origin
Extracted from Acquired. Yakovenko frames the moment as noticing that a root certificate and trust-on-first-use already underpin the internet, so accepting weak subjectivity was not a novel concession — and if you bite that apple, you may as well farm the orchard.
Core principles
- 01Half-paying for a property gets you the cost without the benefit.
- 02Name the axiom you are abandoning before you design around it.
- 03Precedent matters: check whether the wider world already accepts the weaker assumption.
- 04The competitor stuck in the middle is optimizing for two contradictory promises.
- 05Full commitment turns a compromise into a design space nobody else is exploring.
How to run it
- 1
Name the guarantee you are giving up
Say it plainly in one sentence. Here: a new participant cannot verify the chain objectively from first principles and must first establish which network is the right one.
Pro tip If you cannot state it crisply, you have not decided yet — you are drifting.
- 2
Check that the cost is already sunk
Establish whether partial compliance actually preserves anything. Weak subjectivity is binary: you are either objectively verifiable or you are not, so hedging buys nothing.
Pro tip Look for the point where the property becomes all-or-nothing — that is your permission to commit.
Watch out Some guarantees genuinely degrade gracefully; do not force this framing where it does not fit.
- 3
Find the precedent that normalizes it
Locate an existing, widely trusted system running on the same weaker assumption. Root certificates and trust-on-first-use make the concession ordinary rather than reckless.
Pro tip A precedent is also your best answer to the objection you will hear a hundred times.
- 4
Inventory the inherited constraints
List everything in your design that exists only because the old guarantee did. Growth caps on data, conservative throughput, and verification-friendly simplicity all belong to proof of work's promise.
Pro tip Ask of each constraint: who exactly is protected by this now?
- 5
Delete them and re-derive the goal
With the constraints gone, restate what you are optimizing. The question stops being 'is this sound money' and becomes 'what is the most censorship-resistant, fastest messaging network we can build'.
Pro tip Write the new objective as one sentence and test every later decision against it.
Watch out Reintroducing an old constraint under a new name is the most common way this fails.
- 6
Say out loud what you are not
Declare the promise you are no longer making. Solana is explicit that it is not competing to be digital gold or sovereign money, which frees it from arguments it would otherwise have to win.
Pro tip The disclaimer attracts exactly the builders who did not want the old promise anyway.
In the wild
Rather than positioning as a better store of value, Yakovenko sets store-of-value aside entirely. Bitcoin's claim rests on thermodynamic verification of every state transition back to day zero, which caps how fast the chain can grow. Having already accepted weak subjectivity, Solana treats the throughput cap as optional and aims instead at a censorship-resistant messaging and marketplace layer — the analogy being Nasdaq handing every participant an identical-length Ethernet cable so nobody can get ahead.
→ A design free to chase throughput, and a positioning that sidesteps the store-of-value fight altogether.
An illustrative counter-case tied to the same mechanism: a platform team adopts an eventually-consistent datastore for scale but keeps synchronous cross-region confirmation on every write 'to be safe'. They have already lost strict consistency guarantees, yet pay full latency for them. Naming the forfeited guarantee and removing the leftover synchronous step is the whole of the fix.
→ Latency collapses toward the theoretical floor and the system finally behaves like the model it actually chose.
Common mistakes
Hedging after the trade is made
Keeping partial constraints from the abandoned model preserves none of its guarantee while surrendering the performance the trade was supposed to buy.
Committing without a precedent
Going all in on a weaker assumption with no widely trusted system running the same model turns a considered trade-off into an unforced credibility problem.
Still selling the old promise
Marketing the property you deliberately gave up invites the one comparison you cannot win and confuses the builders your new design would actually suit.
Is it for you?
Best for
Architects and founders entering a category where the incumbents share one unexamined constraint.
Not ideal for
Situations where the abandoned guarantee is the actual product promise customers are buying.
From the transcript
“And once you kind of bite that like apple of weak subjectivity, you might as well go all the way”
“trust and first use is a standard security model”
“If you go all in, then you can just start optimizing this thing to the limit.”
From the episode
Special: Solana (with CEO Anatoly Yakovenko)