Shipped a highly requested update to OnchainInt.
Every smart address trade now shows their full position. Bought amount, realized + unrealized P&L, % still holding.
You can also filter by smart addresses and/or KOLs.
Enjoy ๐
This may be a little weird, but I think about this post a lot.
Marketing is hard. Founder-led marketing especially. You have to constantly pump out authentic, engaging content - taking a lot of time & effort. All to convince peopel to use your product. But the ROI is much much much higher.
Snooper runs a few evergreen websites (non-crypto). Fully SEO'd and I get free dofollow high quality backlinks all the time - even referenced by Wikipedia and many Universities. The websites average about 300k monthly visitors.
And I still donโt know how to get 100 people to try a product I built. A product I've made massive amounts of money with.
SEO is a skill. Founder-led marketing is a different sport.
My goal this year is to master it.
Crypto keeps losing 9 figures to problems airplanes and trains solved 50 years ago.
Every safety critical system follows one rule: no single failure at any single step.
The design principles are strictly enforced:
- Defense in depth
- Each step gets its own layer of redundancy
- Each layer is built as if every other layer has already failed
- Assume the worst
Redundancy inside a layer is voting. Read "1oo1" as "one out of one."
- 1oo1: one voter, must say yes. Zero fault tolerance. Compromise the voter, the layer lies.
- 2oo2: two voters, both must agree. Safer, but one going offline halts the layer.
- 2oo3: three voters, any two must agree. Industrial default. One voter can be compromised, offline, or wrong and the layer still answers correctly.
Redundancy only works when the failures are independent.
Real redundancy needs diversity: different implementations, different data sources, different physical locations, different operators. Everything different. On top of that, third party pen tests and audits run strict and frequent. Audits verify the architecture is followed & enforced.
LayerZero, just on the RPC side, should run 3oo3 with one node required to be operated internally. Every incoming message checked against the other two. A 1 of 1 DVN config should never have been permitted in the first place.
$290M moved through a 1oo1 system ๐
Remember, you are only as good as your weakest link.
AI gives me the same excitement as crypto does.
I can build anything in a day.
What I run right now as one person snooper:
- SEO content creation ranking directory sites across multiple niches
- automated tiktok + instagram posts every day for direct customer marketing
- google maps scraper pulling leads by niche, city, and rating
- cold DM sequences going out to every lead automatically
- a live crypto terminal tracking smart money wallets in real time
- personalized outreach where I build MVPs for individual businesses. "if you like it, you can buy it"
Now is the greatest time to build. No team. No funding. Just you and an AI subscription.
๐ I read the full blog post behind this. Some of the wildest AI research I've ever seen.
AI Improvements:
Previous Claude wrote 2 working exploits out of hundreds of attempts. Mythos wrote 181. Same test. 90x improvement in one generation.
What it found:
- 27 year old bug in OpenBSD.
- 16 year old FFmpeg vulnerability every fuzzer on earth missed.
- 17 year-old FreeBSD root exploit.
- Zero-days in every major OS and browser.
- Weaknesses in the world's most popular cryptography libraries: TLS, AES-GCM, SSH.
What it built:
- A browser exploit chaining 4 vulnerabilities into a JIT heap spray that escapes the sandbox and writes directly to the kernel...From a webpage ๐
- Linux kernel root by independently finding and chaining 3-4 separate vulnerabilities.
- A FreeBSD chain split across 6 network packets for unauthenticated root from anywhere on the internet. wtf??
The best part:
They didn't train it for this. The hacking ability emerged as a side effect of general improvements in code and reasoning.
Engineers with zero security background asked it to find vulnerabilities before bed. Woke up to a working remote code execution. Cost per bug: under $50.
Anthropic won't release it. 40 partners only: Apple, Google, Microsoft, AWS, CrowdStrike, JPMorgan, Nvidia. $100M in credits. $4M for opensource security orgs.
99% of what they found is still unpatched. They published SHA-3 hashes as proof they'll disclose everything once it's fixed.
Absolutely incredible
Excited to finally start this hackathon!
> Building a real time crypto trading tool that shows you what high-performing wallets are buying and selling on Solana.
> All based on my profitable trading system
Get ready for the surprises ๐
Does anyone have a list of good EVM contract audits?
Looking for some free/automatic initially before I pay for a real audit.
I already used SolidityGuard by @yq_acc and it is absolutely incredible. But I want a couple more.
Here are the top trades from @OnchainInt over the past 7 days, including memescope monday.
Interesting facts:
- most of our smart addresses were idle during this time
- majority of these winning trades were quick flips
- the longest hold had the largest ROI: within an hour, the top trade had about 25 sells to close out the position.
- even with memescope Monday, we are still in a bear market ๐
๐
Blows my mind how just 7 months ago I was paying $200/month to use Claude Sonnet 4.5 for programming.
Now, Google's Gemma 4 model can be run on my local hardware for free. Achieving almost idential performance.
๐คฏ
The part Google left out of the tweet:
> Gemma 4 comes in 4 sizes. The biggest (31B) just ranked #3 in the entire world on Arena AI. That's where real users vote on which AI gives better answers. Out of every open model on earth. Third.
The 26B model is even more interesting.
It uses something called "Mixture of Experts."
In English: 26 billion parameters total, but only 4 billion activate per task. Like having 26 employees but only 4 show up to any meeting. That's why it runs on a regular laptop despite being massive.
How smart is it?
89.2% on AIME 2026. AIME is one of the hardest math competitions on the planet. Most humans fail it. This model nearly aced it.
Codeforces ELO of 2,150. That's competitive programmer territory. It writes code better than 95% of devs on the platform ๐
256K context window = it reads and understands roughly 200 pages of text at once.
The smallest version (2.3B effective) runs on a Raspberry Pi. The 26B runs in your browser. No server. No cloud. Your browser tab, via WebGPU.
And the license change matters more than the benchmarks.
Previous Gemma versions had usage restrictions on what you could build. Gemma 4 ships under Apache 2.0 for the first time. Build anything. Ship commercially. Modify the weights. Zero strings attached.
Google just gave away one of the best AI models. For free.
๐ฐ The smart address, divergent-aardwark, bought $175 worth of CHIBELON and sold for a profit of $3,450. Almost perfect execution.
โฒ๏ธ Total position time was 50 minutes
๐ We track this wallet live on @OnchainInt
Here's what I've been doing:
1. Contacting friends & acquaintances who have been very successful in their niche/profession.
2. I get a brain dump of literally everything. Mainly focusing on current solutions, pain points, customer acquisition methods, market size, opportunities, etc.
3. Using the input, I create an MVP, utilizing AI that solves the main paint point. It's a polished product that does one thing INCREDIBLY well.
4. The person from step 1 then works on the distribution to acquire customers. The customers are literally their colleagues from over the years.
๐
smart addresses we track bought $UHY and flipped it for a ~4x in about 6 hours.
the best feeling as a founder is when your users make money from something you built
Bro has 90.7% win rate strictly trading pumpfun tokens these past 2 months. Massive earnings.
Tell me you're part of a cabal without telling me you're part of a cabal.
This isn't even the most impressive one either ๐