What a Pump.fun sniper bot is actually buying
On Pump.fun a token does not start life in a liquidity pool. It starts on a bonding curve, which is a program holding SOL on one side and tokens on the other, quoting a price to anyone who asks. You won't find a market maker, an order book or anybody on the other side of your trade, because the program itself is the market and it quotes to everyone on the same terms.
Every launch has the same shape: one billion tokens, a curve that sells them at a rising price, and a threshold that ends the curve and moves everything into a real pool. Buying at second three means buying from a program rather than a person, and that changes what can go wrong.
Hold on to one idea before anything else: risks on a curve are human risks. Nobody drains a pool that doesn't exist yet.
Why a general sniper bot struggles here
- It watches pool creation, and on a curve there's no pool to watch, so the whole launch happens invisibly and the bot turns up at graduation with everyone else.
- It screens for contract traps, but Pump.fun removes those by design, so the screener returns a clean result on every single launch and quietly stops being a filter.
- It sizes for pool depth, and a 0.5 SOL buy into an empty curve behaves nothing like the same buy into a pool holding 400 SOL, which is how people end up filled 40% above the quote and blame the bot.
How the bonding curve prices your buy
Underneath, a Pump.fun curve is a constant product market maker with virtual reserves. Two numbers multiplied together always give the same result. SOL goes into one side, tokens come out of the other, and since that product has to stay constant, every token you buy costs more than the one before it.
Your own buy moves the price against you before it lands. That one sentence is the whole reason curve sniping is a different job from pool sniping.
| Where the curve is | What 0.5 SOL does | What it means for you |
|---|---|---|
| First seconds, almost empty | Moves the price a long way in one transaction | You need wide slippage or the buy simply fails |
| Filling, a few dozen SOL in | Moves it noticeably but not violently | The window where most snipes are actually placed |
| Near the graduation threshold | Barely moves it | Cheap to enter, but the easy multiple is already gone |
| After graduation, in a pool | Depends on pool depth, not on the curve | Normal AMM behaviour, tighter slippage works again |
So a fixed slippage number is wrong somewhere on every launch. Fifteen percent that works comfortably at minute ten will reject your buy at second two, and thirty percent that lands at second two will overpay badly at minute ten, which is why a bot that treats where the curve sits as an input rather than a constant fills more often and pays less doing it.
What happens in the first slot
Five things happen in order every single time, and it's over faster than a page can load.
- 1The create instruction landsOne transaction creates the mint, the metadata and the curve account. This is the moment a bot has to see, because everything after it is already a race.
- 2The creator may buy in the same transactionPump.fun lets the creator take a position as part of creation. That buy is priced at the very bottom of the curve, which is exactly where you would want to be.
- 3Bundled buys land in the same slotA group can submit several buys atomically through a Jito bundle, so a dozen wallets are already holding before the mint appears anywhere public.
- 4The first outside buys arriveThis is where a sniper bot competes. Not against the creator, who was earlier, but against every other bot reading the same logs.
- 5The chart appearsBy the time a token shows up on a screener feed, the first four steps are history and the price already reflects them.
None of it is hidden. All of it sits on chain, readable in the same slot it happens. Whether a bot profits or quietly donates comes down to one thing: does it read step two and step three before committing to step four, or does it fire because something new appeared?
What to screen when the authorities are already null
One fact changes how you screen Pump.fun, and nearly every thread on the subject gets it backwards. Pump.fun revokes mint authority and freeze authority when it creates the token. Both are null on every launch.
So the two checks everyone quotes pass on every token, including the ones that go to zero in four minutes. They aren't protecting you. They're answering a question nobody asked.
No LP token exists either. Liquidity on the curve belongs to the program, so a burn check has nothing to inspect until the token graduates, and that leaves three of the most quoted safety checks in Solana trading with nothing to say about the venue where most Solana launches actually happen.
What actually predicts a bad Pump.fun launch is behavioural, and all of it is measurable:
| Signal | What it reads | Why it predicts a loss |
|---|---|---|
| Creator holding | The share of supply the deployer took at creation | Someone holding 20% of a curve is a seller waiting for your buy to give them a price |
| Bundled opening | Buys landing in the same slot as the create instruction | The supply is already placed, and every later buyer is exit liquidity for that slot |
| Funding graph | Where the earliest buyer wallets got their SOL | Twenty wallets funded from one source are one seller wearing twenty hats |
| Deployer history | Earlier mints signed by the same wallet and how each ended | A wallet that abandoned four launches last week is the single most predictive signal available |
| Metadata quality | Whether the URI resolves and the socials exist | A mint with broken metadata rarely has anyone intending to build anything |
None of that needs a paid data provider. It's account state and transaction history, readable from a node in the same slot the mint appears, so a bot can check it before it buys rather than after you complain.
What changes when the curve fills
Graduation is the moment a curve hits its threshold and stops being the market. Collected SOL and leftover tokens move into a real pool, and from that second the token behaves like any other Solana pair.
Five things flip at once, and a bot that carries its curve settings through the handover gets all of them wrong.
| On the curve | After graduation | |
|---|---|---|
| Who quotes the price | The bonding curve program | A liquidity pool with real depth |
| Slippage that works | 15% to 25%, sometimes more | 10% is usually plenty |
| LP checks | Nothing to check, no LP exists | Burn, lock depth and unlock date all matter |
| Who can pull liquidity | Nobody, it is program owned | Whoever holds the LP tokens, if they were not burned |
| What kills the position | The creator selling their allocation | Liquidity leaving the pool |
Our engine re-reads the token at the handover, so an open position carries across, the pool level checks start running against the new pool, and your exit rules stay armed the whole way through. If the LP turns out to be sitting in the deployer's wallet rather than burned, the auto rug exit is already watching it.
Most of the money in a Pump.fun trade is made or lost in the minutes either side of that moment. Watch only the launchpad and you see the entry but miss the exit. Watch only the AMM and you turn up after the part that mattered.
Settings that only make sense on a curve
Moving from pool sniping to curve sniping? These are the numbers we'd change first.
| Setting | Fresh curve | Graduated pool | Why they differ |
|---|---|---|---|
| Buy slippage | 15% to 25% | 10% | Your own order moves an almost empty curve and barely moves a deep pool |
| Buy size | 0.1 to 0.25 SOL | 0.25 to 0.5 SOL | Smaller size on a curve buys a better average price, not just less risk |
| Let the coin age | 0 to 3 seconds | 0 seconds | A few seconds on a curve lets the bundle detection and holder read finish |
| Take profit | 60% to 150% | 40% to 80% | Curves move further and faster in both directions |
| Maximum hold | 15 to 30 minutes | 30 to 60 minutes | A curve launch that has not moved in twenty minutes usually never will |
| Sell slippage | 25% to 40% | 20% | Getting out of a thin curve costs more than getting in |
One number deserves its own paragraph. Setting the age filter to zero buys you the very first block, and it buys you every bundled launch along with it, because nothing had time to spot the bundle. Two or three seconds is often the whole difference between a clean fill and being the exit for a group that was already inside.
LetsBonk, Bags, Boop and the rest
Pump.fun proved the model and everyone else copied the shape: a curve, a fixed supply, a threshold, a pool at the end. Underneath, each one differs in its program, its fee split, where it graduates to and how its instructions are laid out.
That last part is why coverage is real work rather than a checkbox. Every launchpad emits its own instructions and its own account layout, so claiming ten launchpads means decoding ten different programs, and a bot that reads one and hopes the rest look similar will miss launches quietly and blame the network when you ask.
| Launchpad | What is different | Graduates into |
|---|---|---|
| Pump.fun | The original curve, deepest flow | PumpSwap |
| LetsBonk | Same curve model, different fee split | Raydium LaunchLab |
| Bags | Creator fee sharing built into the launch | A pool at graduation |
| Believe | App and internet capital market launches | A pool at graduation |
| Boop.fun | Curve launches with staking attached | A pool at graduation |
| LaunchLab | Raydium's own launch curves | Raydium |
| Jupiter Studio | Studio mints with locked vesting | Jupiter routed pools |
| Heaven | Curve launches with a buyback mechanic | A pool at graduation |
| Daos.fun | Fund raises rather than meme launches | A pool at graduation |
| Vector | Launches driven by a social feed | A pool at graduation |
Each of those venues has its own program, threshold and fee model. The pages below cover the five that matter most:
- LetsBonk sniper botBonk.fun on LaunchLab rails, graduating near 85 SOL
- LaunchLab sniper botRaydium's configurable curves, read per launch
- Bags sniper botMeteora curves and a royalty that never stops
- Believe sniper botthe venue that charges you for arriving first
- Boop.fun sniper bota 400 SOL curve and a 5% airdrop at graduation
Five of them are on the 1 SOL Scout plan: Pump.fun, LetsBonk, Bags, Boop.fun and LaunchLab. The 3 SOL plan adds the other 5 along with 14 more AMMs, books and routers, and the 5 SOL plan opens all 42 venues.
What curve snipers get wrong
- 1Trusting the null authoritiesMint and freeze authority are revoked on every Pump.fun launch. Seeing both green tells you the launchpad works, not that the token is worth buying.
- 2Using pool slippage on a curveTen percent into an empty curve rejects most of the time, and every rejection still costs the base fee plus your priority bid.
- 3Buying in slot zero on purposeZero delay means zero time to detect a bundled opening. You arrive first and you arrive as the counterparty for wallets that were already inside.
- 4Keeping curve settings after graduationTwenty five percent slippage into a pool holding four hundred SOL is a donation. The trade changed character and the numbers should change with it.
- 5Ignoring the creator's other mintsDeployer history is the strongest predictor on this venue and it takes one lookup. A wallet with four abandoned launches behind it will not behave differently on the fifth.
Questions people ask
These come up in the ticket queue often enough to answer once, properly. If yours is not here, send it over and it probably ends up on this page.
Technically yes, and it is usually the wrong goal. A transaction built from the create instruction can land in the same slot, but so can the creator's own buy and any bundle they submitted with it. Landing there means buying alongside the people who set the launch up, with no time to have read anything about them.
Key takeaways
- Pump.fun tokens trade on a bonding curve, so there is no pool, no LP token and no liquidity provider until graduation.
- Mint and freeze authority are revoked on every launch, which makes the two most quoted safety checks useless on this venue.
- What predicts a bad curve launch is behavioural: creator holding, bundled opening, funding graph and deployer history.
- Your own buy moves an empty curve, so slippage has to be wider early and tighter later.
- Graduation flips the trade into ordinary AMM behaviour, and settings that worked on the curve stop working there.
- Waiting two or three seconds usually beats buying in slot zero, because it leaves time to detect a bundled launch.