you asked. paper trading is coming back.

Parasol agent mid-velocity scalp — the lab is where you watch this happen safely before going live

in march, we discontinued public paper trading on parasol. it was the right call at the time. it stung anyway.

a lot of you told us, in dms, on twitter, in our community channels, in the open beta feedback forms, that you missed it. you used it to test a strategy without risking real sol. you used it to learn how the dashboard worked before connecting a wallet. you used it to compare agent personas head-to-head. some of you used it to convince yourselves the engine was real before going live.

we listened. and we worked out a way to bring it back without the cost problem that killed it the first time.

the performance lab is now restored at parasol.so/dashboard/performance, behind a feature flag, open to testers this week. here's what we did and why.

---

why we ended it the first time

the original paper trading retrospective goes deep on the data — 17,723 trades over 23 days, the strategies that worked, the strategies that didn't, the bug that was cutting winners and leaving real money on the table.

the part of the retrospective that mattered for the decision to discontinue was about cost. a paper trade is, from an infrastructure perspective, indistinguishable from a live trade. it makes the same calls to the same upstream providers. it pulls the same price data. it runs the same scoring. the only thing that changes between paper and live is that the swap at the end is simulated rather than executed on-chain.

we had 24 different agent modes running in parallel during the public paper trading window. tens of thousands of upstream calls per hour. the cost of running paper trading at that scale exceeded the revenue we collected from live trading it was meant to validate.

so we paused it. the code stayed in the repo, fully preserved. we knew this wasn't the right permanent answer. we just needed time to design a permanent answer that didn't burn down our infrastructure budget.

---

what's different this time

phase a: browser-side first.

the version that's open to testers this week runs the parallel paper-trading engine in your own browser session. the upstream calls happen from your machine, not from our infrastructure. that means the cost of running it scales with the user, not with our hosting bill. one user testing a setup is, infrastructure-wise, almost free for us.

phase a is browser-side, with no per-minute billing. it's gated behind a feature flag (NEXT_PUBLIC_LAB_ENABLED) so we can roll it out to testers gradually before opening it more widely.

phase b: hosted, metered, opt-in.

the next sprint moves the runner to our own hosted infrastructure, the same one running 24/7 server agents, with per-minute pcu billing on top of the existing credit ledger. paper trading becomes a metered tier of the parasol product instead of a free-to-everyone external api drain. you'll pay a small amount of pcu for every minute your lab runs. the same pcu that powers metered mcp calls, the same balance, the same top-up flow.

that's the lesson we should have learned the first time. the original retrospective said the right thing but at the wrong granularity. the lesson wasn't "paper trading is bad." it was "paper trading without a usage model is bad." with metered usage, the unit economics work for everyone.

---

Mei in deep zen — the calm of running a strategy in the lab before risking real sol

what the lab gives you

the performance lab is the same parallel engine that ran 17,723 trades during the original window. you can:

run multiple agent personas in parallel against live market data

compare their performance side by side over the same time window

test risk profiles (careful, aggressive, degen) without risking real sol

evaluate strategy modifications before pushing them to your live agent

get a feel for how the engine behaves in different market regimes before committing capital

every paper trade captures the same feature snapshot a live trade does. the same scoring runs. the same exit logic fires. the only difference is the swap at the end.

if you've used parasol live and wondered "what would have happened if i'd had grok running too" — that's exactly the question the lab answers.

---

who can get access this week

if you're an existing parasol user and want to be in the first batch of phase-a testers, the easiest path is:

1. dm @parasolsolana with your account email or wallet address

2. we add you to the testers list and flip the flag for your account

3. you go to /dashboard/performance and start running setups

we'll roll out gradually. we want feedback at each stage rather than opening it to everyone at once and missing edge cases.

if you're not yet a parasol user and you're interested in the lab specifically, parasol.so/beta is the way in. mention "performance lab" in the note field and we'll prioritise the access code.

---

what we'd like to learn

three things specifically, from this round:

1. does the browser-side runner perform well enough? running upstream calls from a single browser session has different latency characteristics than running them from a server with a fast direct connection. we expect a small degradation in scan cadence. we want to know if it's acceptable in practice or if it pushes phase b's hosted version up the priority list.

2. what setups do users actually want to test? persona vs persona is the obvious one. risk profile vs risk profile is the next. but the most valuable test we ever ran ourselves was strategy-by-market-regime — and we want to know whether users care about that, or whether the comparisons that matter to traders are different from the ones that matter to us as engineers.

3. what would phase-b pricing look like that you'd actually pay? per-minute billing has the right unit economics for us. the question is what level of pcu-per-minute matches the value users get from running a 4-hour comparison test. dm us with your honest answer.

---

Ghost — a quiet position taken without anyone noticing, the lab teaches you how

a small note on owning a mistake

it's worth saying out loud: ending paper trading the first time hurt people we built parasol for. the cost rationale was real but the comms were minimal and the gap between "ending" and "bringing back" was longer than it should have been. we're sorry for the gap.

we kept all the code. we knew this wasn't permanent. it just took us longer than expected to find a usage model that lets us bring it back without re-creating the original problem.

phase a is live now for testers. phase b in the next sprint. by the end of may, paper trading is back as a real, sustainable feature with a clear path between testing a setup and running it live with confidence.

if you used paper trading before and missed it, this is your moment.

— the parasol team