most retail traders, asked what a market maker does, will say "they manipulate the price" or "they scalp spreads from people like me." this is wrong in a way that matters. market makers are, in aggregate, one of the few pieces of financial infrastructure where the private interest and the public interest are aligned almost perfectly. a good market maker makes tight quotes available continuously, absorbs the cost of adverse selection, and produces the thing every trader actually needs: a price you can trade at right now. the reason retail thinks of them as adversaries is that the tradfi market-making business is dominated by a few firms with huge technology advantages, and retail's only exposure to the industry is losing money to those firms. the response should not be to make market-making harder. it should be to make it accessible.
that is the thesis underneath hyperflash, which is one of the products i have been working on. the short version: institutional-grade market-making and grid-trading strategies, deployable without code, running on hyperliquid and a growing set of perp dexes. the slightly longer version is a story about how crypto perp infrastructure has matured to the point where the sophisticated end of automated trading is now a commodity — and why commoditizing it is a public good.
what market-making actually is. a market maker posts two-sided quotes (a bid and an ask) around a reference price, collects the spread when both get hit, and inventories the residual when order flow is imbalanced. the profit comes from the spread. the risk comes from informed traders who pick off stale quotes faster than the maker can update. in a normal venue, the maker's edge is speed — quoting faster than informed order flow can detect stale prices — and inventory management — not accumulating so large a position in one direction that a regime shift blows up the p&l.
in practice, running a decent market-making strategy on a crypto perp dex requires: a streaming connection to the venue, risk controls to cap inventory per asset, fill-aware pricing that widens when the maker is getting picked off, a hedging leg for delta-neutral variants, and monitoring that detects when the venue has stopped behaving normally. none of this is rocket science. all of it is annoying to build from scratch, and the tail risks — a single bug, a network partition, a flash move — are severe enough that most retail never deploys it in production even when they could.
why the platform matters. when you look at tradfi, the market-making edge at the top end is not primarily strategy. the strategies have been public for thirty years. the edge is infrastructure — colocation, order-routing optimizations, risk systems, kill switches, the thousand small engineering decisions between a backtest and a live book. a retail trader in tradfi cannot replicate this for perhaps ten million dollars of annual cost. in crypto, the same infrastructure can be built as a shared platform and rented to retail users for a fraction of that. the cost of running ten thousand retail market makers through shared infrastructure is not much higher than running one, which means the per-user cost can approach the marginal cost of quotes rather than the fixed cost of infrastructure.
a platform like hyperflash is mostly a set of opinionated defaults and a few good primitives: a market-making bot with sensible spread and inventory logic, a grid bot for range-bound assets, a delta-neutral bot that hedges spot with funding-capture perps. the user picks the venue, the asset, the risk envelope, and the bot runs. the mechanism design is in the defaults. bad defaults produce blowups. good defaults produce an asymptotic approximation of a professional market-maker's p&l with retail-sized capital.
why this is a public good. every additional market maker tightens spreads and improves price discovery for every other user of the venue. the benefit of a tighter spread accrues to everyone who trades that asset — takers, other makers, casual speculators, even passive holders whose mark-to-market is cleaner. but the cost of providing tight spreads accrues privately to the maker who is adverse-selected. this is the standard public-goods asymmetry: marginal benefit diffuse, marginal cost concentrated. you would expect undersupply, and in crypto you get it — venues outside btc and eth have visibly wider spreads, thinner books, and more slippage on basic order sizes than they need to.
the platform intervention is a classic public-goods fix: reduce the private cost of participation until the marginal maker shows up. if a retail user can spin up a market-making bot in five minutes, pay a flat fee, and run it on hyperliquid, the supply curve of liquidity shifts right and the long tail of perp assets becomes tradeable. everyone benefits. the maker captures enough spread to make it worth their time. the venue gets better execution quality. the next taker pays less slippage. the equilibrium shifts.
what i have been surprised by. three things.
one: naive grid bots work better than they should on assets with mean-reverting funding. the textbook argument says grids lose money in a trend. in practice, most perp assets outside the top few are dominated by range-bound mean reversion punctuated by occasional shocks, and a well-parameterized grid extracts meaningful yield. the shock risk is real and has to be capped, but the base rate is favorable.
two: delta-neutral funding capture is the single most underutilized strategy in retail crypto. long spot, short perp, collect funding. the risk is basis risk and staking unavailability, which can be managed. the return is often better than any yield-bearing product a cefi exchange will offer, and it is fully transparent. the reason it is underutilized is that building it yourself requires coordinating two venues and monitoring funding schedules, which is exactly the kind of friction a platform dissolves.
three: the single biggest source of user blowups is not bad strategy; it is good strategy with a parameter set for a different regime. a market-making bot that was printing money in q2 can hand back the gains in q4 if the user does not widen spreads as volatility rises. the platform's job is to default those adjustments, and to refuse to let the user run strategies that are obviously underpriced for current regime.
building hyperflash is mostly microstructure work — the economics is clean, the strategies are known, the crypto rails are finally fast enough. the part i find most interesting is the mechanism-design layer on top: how do you set defaults that protect a retail user from themselves without making the product so safe that it has no edge? that is the question i expect to keep answering for the next year.
the docs, if you want to poke at the mechanics in more detail, are at docs.hyperflash.xyz.