The Shittttcoiner retweeted
I trade on Lighter too, and after today’s incident i’m going to be a lot pickier about what i give an API key to..
Lighter says unofficial signing software exposed a trader’s API keys. the contracts weren’t breached, but $10.4m still left the exchange. 5M of which i think is recovered now..
so for my own setup:
- if a bot can’t tell me whether it uses Lighter’s official SDK to sign orders, i’m not connecting it
- if i can’t verify how an old key was used, i’ll revoke it and start with a fresh one
- i’ll run bots on a separate subaccount with only the funds they need
Our team has been hard at work to help a customer recover 5MM USDC when their API keys were leaked and exploited by an attacker. We recently launched the Guardian service and it was key to being able to detect the anomaly and resolve in a way consistent with the verifiable architecture of Lighter.
Importantly, the escape hatch provides a safety net in the event of a protocol-wide attack (which this was not, thankfully). As we continue to invest in security, we strongly encourage traders on Lighter to follow best practices for their own security setup, including API keys, as we refer to in the report below.
xitter.gq/Lighter_xyz/status/212…
The Shittttcoiner retweeted
Stocks are bouncing hard today even with NVIDIA getting hit..
we were worried that a selloff in $NVDA would drag the entire market down with it.
so far, the opposite is happening: investors are buying other tech stocks while NVIDIA deals with its own headlines.
oil is falling too, which could help ease some of the inflation fears that have been weighing on markets.
now the next big test is tomorrow’s US inflation data. if that comes in better than expected, i think it could give this recovery more room to run.
and for crypto, i’m watching if BTC can hold up while stocks recover. would be nice to finally see some of that confidence come back here too.
The Shittttcoiner retweeted
Give me a one good reason not to fade ARC ?
I don’t see a catalyst that brings fresh money into its memes once the launch hype fades..
- Robinhood has Vlad openly supporting the ecosystem and listing coins doing good on chain giving traders something to look forward to.
- Base chain has Coinbase distribution and now Cobie at the frontend
- Solana already has the meme traders, liquidity and communities. i can see why money would go there..
- And ARC atm feels like launchpad wars.
- i’m struggling to find a meme building a cult beyond people waiting for the next launch. the whole thing feels forced.
- My test is whether one of these launchpad coins reaches $100m mcap in the coming days. if none can, that strengthens my view that demand isn’t there. a brief wick wouldn’t convince me either..
so i think the next rotation could go back to solana memes or robinhood DeFi.
that’s where i’d rather look until ARC gives me a reason to change my mind.
The Shittttcoiner retweeted
My screening setup to find launches to ape on ARC chain:
there are already more coins than i can follow, so i want to narrow things down before spending time on individual charts.
i keep the Arc page open on GeckoTerminal with:
- 30K+ liquidity, to cut out the thinnest pools
- 50K+ daily volume, so there’s some trading activity
- 100+ holders, to exclude coins with barely any wallets
- top 10 holders at 40% max combined, to limit obvious concentration
the idea here is simple.. i’m willing to miss some first-minute entries if it means having more information before buying.
then i compare the 1h and 6h moves with the 24h. a coin being up 500% today doesn’t tell me whether it’s still moving or has been dumping for the last three hours.
if i like the chart, i open Transactions and Holders to check the trade sizes and who controls the supply.
the filters give me candidates, and these checks help me decide which ones deserve more time..
also, i check “Other Pools.” the same token can have several pools with different liquidity and fees, so i compare them before checking the actual quote in my trading app.
that’s how this helps me.. fewer random charts, the same checks for every coin, and less rushing into something just because it landed on my timeline.
The Shittttcoiner retweeted
Just in case this ends up mattering as eligibility criteria:
> Download the Pumpfun app
> Make a few trades (stick to higher mcap coins to avoid getting killed on slippage)
> Create a coin
> Make a few posts on the App
> Follow some accounts… you can follow me AxelBitblaze too
The Shittttcoiner retweeted
i’ve never done a giveaway before, but if i manage to turn this into $100K,
i’ll take $5,000 and send it equally among everyone who comments on the post below before i hit $100K
please God, let me come back to this post.
The Shittttcoiner retweeted
NFT mints are printing free money right now.. these are the tools im using to find upcoming mints early to stay ahead:
1) Grok: scan what NFT traders are following and talking about
2)@alphabot : find new WL raffles and project collabs
3) @nftcalendar : see what’s minting over the next few days
4) OpenSea Drops: check confirmed mints and my wallet eligibility
5) Magic Eden Mint Terminal: find what’s upcoming or already minting across different chains
So yeah Grok basically finds them early for me and the rest help me track the mint..
Don’t forget to bookmark this
The Shittttcoiner retweeted
I just built this AI integrated portfolio tracker for myself..
and bro… this is just next level, it shows me:
- My drawdown
- Portfolio volatility
- Sharpe ratio
- And whether my returns are even worth the risk i’m taking
i can also ask the AI questions about my holdings.. and run Monte Carlo simulations to see where it could realistically end up across thousands of possible market paths.
The Shittttcoiner retweeted
lowkey glad we went through such a brutal bear market since 10/10
i was kinda left with no option but to learn something new.. and so glad i chose AI
now i’ve been building cool shit ever since. like u have seen this Ai integrated portfolio tracker..
been using AI to track smart wallets, scan live market data, test my theses, find crowded trades, monitor risk and automate parts of my execution..
love it
I just built this AI integrated portfolio tracker for myself..
and bro… this is just next level, it shows me:
- My drawdown
- Portfolio volatility
- Sharpe ratio
- And whether my returns are even worth the risk i’m taking
i can also ask the AI questions about my holdings.. and run Monte Carlo simulations to see where it could realistically end up across thousands of possible market paths.
The Shittttcoiner retweeted
#5 Task: @nansen_ai x Claude
Let’s use Claude + Nansen to audit any Solana trader before you copy their wallet.
paste a “top trader” from FOMO or Pumpfun app.. and Claude will check whether their profits are actually consistent and copyable..
or whether the leaderboard is hiding one lucky trade, unsold bags or tokens they never bought themselves..
by the end, we’ll know:
- realized PnL and win rate
- 7d, 30d and 90d performance
- how much profit came from one token
- PnL without their biggest winner
- current holdings and unrealized bags
- typical position size and holding time
- whether transfers or related wallets distort the result
- whether you could realistically copy their entries and exits
we are not just going to find whether the wallet made money.. but also it’s whether you could have made money copying it.
1/4 🧵 ⬇️
#4 Task: @Helius API
let’s use Claude + Helius to trace who funded the early buyers of the memecoin we’ve been investigating..
because those 20 fresh wallets buying right after launch may not be 20 random traders..
they could all be funded by the same person and made to look independent.
so today, we’re taking the suspicious wallets found during our first 3 tasks and following their $SOL backwards..
Claude will check:
> who originally funded each wallet
> how much SOL it received
> when it was funded
> how quickly it bought the token
> whether multiple wallets lead back to the same source
> whether that source is an exchange, protocol or unknown wallet
5 wallets funded through Binance means nothing but 5 wallets funded by the same unknown wallet minutes before launch is where it starts getting interesting…
by the end, we’ll have a funding map showing whether the early buyers were actually independent..
or just the same operator arriving through different wallets.
1/4 🧵 👇
The Shittttcoiner retweeted
This is already getting lil crazy
I wouldn’t be surprised if soon Grok Bot comes with this update to me:
that he found the job… and the salary offer is waiting for my approval and the company will sponsor my visa to relocate..
until then.. back to work, Scout
I just set up 4 Grok Bots to help me find an AI job abroad..
and they’ve already started passing work between themselves.
Scout checks the career pages of AI companies for roles that match my profile and either allow remote work from my home or mention visa sponsorship.
it just finished its first pass and found 3 Apple AI/ML roles.
those roles are now with Screener, which checks the requirements, location and whether i’m actually a realistic fit.
anything that survives gets passed to Writer… it uses my CV and past work to prepare a tailored application for me to review.
Tracker then keeps the role, application status, replies and follow up date in one place.
all 4 Bots use the same cloud computer, so they can access the same files and continue the handoff even after i close my laptop
gonna leave this running for a week and see if it can bring me an actual interview..
The Shittttcoiner retweeted
#4 Task: @Helius API
let’s use Claude + Helius to trace who funded the early buyers of the memecoin we’ve been investigating..
because those 20 fresh wallets buying right after launch may not be 20 random traders..
they could all be funded by the same person and made to look independent.
so today, we’re taking the suspicious wallets found during our first 3 tasks and following their $SOL backwards..
Claude will check:
> who originally funded each wallet
> how much SOL it received
> when it was funded
> how quickly it bought the token
> whether multiple wallets lead back to the same source
> whether that source is an exchange, protocol or unknown wallet
5 wallets funded through Binance means nothing but 5 wallets funded by the same unknown wallet minutes before launch is where it starts getting interesting…
by the end, we’ll have a funding map showing whether the early buyers were actually independent..
or just the same operator arriving through different wallets.
1/4 🧵 👇
#3 Task: @bubblemaps + Claude
yesterday, we investigated the dev behind the memecoin that survived task 1.
[ DID YOU ACT ON YESTERDAY’S GUIDE ? ]
today, we check whether its biggest holders are actually different people..
fyi, a team can split 20% of the supply across 30 wallets to make holder page look clean..
it may look like no huge whale… just a lot of harmless looking wallets holding 0.5% each..
but if those wallets transferred between each other or connect through one hidden address, Bubblemaps can put them back into the same cluster.
so today we’ll use:
> Magic Nodes to reveal hidden intermediary wallets
> Time Travel to inspect the launch distribution
> Yesterday’s dev wallet to find possible connections
> Claude to compare everything and create the report
no code, API key or wallet connection needed.
by the end, we’ll have day-3-holder-map.md
1/4 🧵 👇
The Shittttcoiner retweeted
i just ran an AI analysis on one of the most profitable traders on FOMO app, @ether_monk…
his trades, entries, exits, patterns, what he does differently, everything..
and this is what it found:
For people using FOMO or Pumpfun app to hunt memecoins this cycle..
do this instead of following a trader just because their pnl looks good..
Open their profile and screenshot their last 20 trades.. make sure the entry, exit, mcap and pnl are visible.
drop them into ChatGPT or Claude with this Prompt:
“ Study these 20 trades and figure out when this trader is actually worth following.
compare their winners and losers by entry mcap, token age, narrative, position size and hold time.
tell me:
- what their best trades have in common
- where their edge actually comes from
- how much pnl came from their biggest trade
- whether someone entering after the alert could still make money
- which future buys i should watch or ignore
when i send a new trade, compare it with their past winners and score it from 0 to 100 based on how well it matches and whether it is still copyable.
only use visible data. mark anything missing as unknown. ”
and when they buy something new, send that trade to the same chat and ask if it looks similar to their previous winners.
if it doesn’t, skip it.
and if AI can’t find any repeatable edge after 20 trades, mute them and move on..
The Shittttcoiner retweeted
#2 Task: @gmgnai Dev Wallet Investigator
yesterday, we gave Claude one Solana CA and built our first memecoin risk report..
( i hope you tried it ? )
now, if your result was REJECT, stop there.
don’t keep changing the prompt until the AI finally tells you what you wanted to hear..
choose another CA and repeat task 1.
if the token survived, today we investigate the wallet behind it..
because a token can look completely clean right now while the dev has:
> launched 30 dead coins before it
> never taken a single one to a DEX
> funded a fresh wallet just before launch
> already started selling or transferring this token
so today, Claude will check for us:
> dev wallet age
> original funding source
> complete token launch history
> graduated vs failed launches
> previous tokens’ ATH and current market cap
> current token balance
> dev buys, sells and transfers
by the end, we’ll have day-2-dev-report.md inside the same research desk.
1/4 🧵 👇
#1 Task: @gmgnai Agent Skills
let’s start day 1 with the part that comes before any entry..
rejecting bad memecoins before even wasting more time researching them…
we all know finding what’s trending is easy.. every scanner can do that, right?
but whats harder is finding what the chart isn’t showing. i mean:
> mint + freeze authority
> holder concentration
> dev holdings
> insiders + bundled wallets
> snipers + fresh wallets
> rug ratio + wash trading
> smart money + KOL activity
Claude’s existing knowledge cannot pull this live onchain data…
so today, we’re connecting GMGN’s official Agent Skills to Claude and building a read-only CA investigator..
by the end, you’ll paste any Solana contract address and receive one of 3 results:
1) REJECT
2) NEEDS DEEPER RESEARCH
3) PASSED FIRST FILTER
“passed” does not mean safe.. and it definitely doesn’t mean buy. it only means the token survived layer one and deserves further research.
complete setup below.
1/4 🧵👇
The Shittttcoiner retweeted
𝕏 just open-sourced the code that decides which posts get pushed and which ones die.
i went through the important parts..
here’s what actually boosts your reach, kills it and quietly hides you from non-followers.
🧵 👇
The Shittttcoiner retweeted
Most of these launchpads solve only one problem:
how do we launch the token?
almost nobody thinks about what happens to the liquidity AFTER it starts trading.
and that's why i find what @TradePools built pretty interesting.
normally on Uniswap, an LP position earns fees as people trade.. but those fees don't automatically become new liquidity.
they accumulate until they're collected and reinvested..
so the pool can generate a shit ton of fees without those fees actually making the market deeper..
but now Pooltrads basically automated that entire loop..
every swap generates LP fees
> a bot sweeps those fees
> sends them through the compounding contract
> puts them straight back into the LP position
> liquidity grows as the market trades
volume isn't just generating fees anymore.. it's continuously feeding liquidity back into the market..
$FRONG is the first really good live example.
i traced 35 compounding cycles so far, with 22.8M FRONG going through that loop.. around $130k at current prices.
and when i checked, uncollected LP fees were basically 0 because the bot keeps sweeping them back in.
this is also why FRONG has had unusually deep liquidity relative to its market cap.
and yeah @haydenzadams has already talked about bringing this same auto-compounding mechanism to regular Uniswap LP positions later…
so imo this is much more interesting than another token launchpad feature.
they're basically making the market itself improve from the activity happening inside it..
simple idea.. but actually a pretty beautiful mechanism.
The Shittttcoiner retweeted
You can run a frontier AI model GLM 5.2 and Kimi K3 on your laptop for free now..
NO NEED TO PAY FOR OPUS 5 and GPT 5.6
no cloud. no API subscription. you don't even need a GPU..
i found this open-source project called Colibrì that basically lets you run GLM-5.2, a massive 744B parameter model, locally.
normally you'd think you'd need some insane GPU setup for that.
but here's the trick..
the model doesn't actually use all 744B parameters at once. only around 40B are active at any moment.
so Colibrì keeps what it needs in RAM and leaves most of the model sitting on your SSD..
when it needs another piece, it pulls it from there.
that's basically how someone got this thing running with:
- 25GB RAM
- 372GB SSD space
- no GPU required
a fast NVMe SSD helps a lot, and obviously a GPU makes it faster.
and setup isn't that crazy either:
1) clone Colibrì
git clone github.com/JustVugg/colibri
2) run the setup:
cd colibri/c
./setup.sh
3) download the recommended GLM-5.2 weights
huggingface.co/mastouri/GLM-…
4) check your laptop
COLI_MODEL=/path/to/model ./coli doctor
5) run it
COLI_MODEL=/path/to/model ./coli chat
now the obvious catch..
25GB RAM means you CAN run it. not that it'll be fast lol.
on the original 25GB machine you're looking at roughly one token every 10-20 seconds.
but the fact that you can run a model this massive locally without owning a GPU at all is pretty wild..
basically you're trading expensive hardware for SSD space + patience.
repo: github.com/JustVugg/colibri
The Shittttcoiner retweeted
> You give Claude Code a long task
> pick up your phone
> 20 mins later you look back at your laptop and it's been sitting on a permission prompt for 19mins of them.
well, this repo fixes that..
When Claude Code needs you, it plays a video at you.. fullscreen, with sound
and it stops the moment you touch the keyboard.
yeah of it checks you're really away first, so it never fires while you're working.
drop your own clips in and get woken by whatever you like.
repo: github.com/rafcopy/claude-co…
The Shittttcoiner retweeted
> every ai you've ever used lives in a tab
> you open it, it wakes up
> you close it, it dies
> mine has a phone number
> it calls ME when something needs doing
> we talk. it goes and handles it. in the real world.
> like checkout the video it even takes trades for me














