writing smart contracts to connect worlds | contributor @hypurrfi
Joined May 2017
- Tweets 17,972
- Following 2,251
- Followers 3,293
- Likes 26,827
1,988 Photos and videos
alphakΞY retweeted
I’ve been through dark times and personal struggles in my life. and as corny as it sounds, this manga panel from hajime no ippo literally saved my life by making me realize how loved i am, and it has continued to inspire within me the will to fight on.
19
4
237
5,767
David Sacks leaving the administration after zeroing the industry & making the Trump family over $1,000,000,000 in crypto
Mar 26
David Sacks says his time as Trump's crypto and AI czar has ended $BTC
What a success.
58
449
9,898
1,297,351
alphakΞY retweeted
Mar 25
btw smart contracts can serve prompt injections
ofc name(), symbol(), etc
but also from behind arbitrary calls if the llm doesn't spec the return type in the cast call :3
---
if an llm runs this against the contract below:
cast call <addr> "owner()(address)"
it'll get `0x20`
---
but if an llm runs this against it:
cast call <addr> "owner()"
it'll get an opaque byte string, which an llm will recognize as a potential string and attempt to decode it
2
6
74
11,739
PSA for LLM enjoyoors
Mar 25
Claude was asked to generate 50 "random" passwords.
Results: only 30 unique. The most common one repeated 18 times (36% probability vs expected 2^-100). All start with uppercase letter 7. Zero repeating characters within any password.
LLMs literally cannot do randomness. Do not ask them for passwords.
3
114
One thing I've learned from using llms is that I'm absolutely right
Mar 24
One thing I've learned from using llms is that I'm absolutely right
3
99
In the course of a couple hours, I built a fully intelligent AI trading agent that combines technical analysis geopolitical news to trade perps on @ostium.
It scores trades based on a TA news weighting that is specific to each asset.
Here's how it works 👇
(only the first 50 to comment will get access)
6
5
208
The TA engine scores across 5 categories:
📈 Trend — EMA 20/50/200, alignment, crosses
⚡ Momentum — RSI, MACD, Stochastic
🎯 Support/Resistance — classic pivot points
📊 Volatility — Bollinger Bands position & width
🔊 Volume — directional confirmation
Each category contributes to a composite score from -100 to 100.
1
1
112
BREAKING: software developers are not losing their jobs.
Personally, I think the current environment is just all companies reacting to what other large companies are doing:
size down their workforce and trying to see what operating looks like with a smaller number of employees.
AWS has already blocked junior and mid-level engineers from pushing AI-assisted code with approval from a senior engineer.
Thus we're already starting to see that large scale AI-assisted software development is hitting bottlenecks when it the AI model chooses to optimize for the wrong thing (e.g. context corruption).
Is now the right time to start a side project with Claude or your open-source model of choice? Absolutely.
Is everyone losing their job in the long-run? Probably not at least until we see another major jump in the models put out by Anthropic and others.
Mar 16
🤯BREAKING: Alibaba just proved that AI Coding isn't taking your job, it's just writing the legacy code that will keep you employed fixing it for the next decade. 🤣
Passing a coding test once is easy. Maintaining that code for 8 months without it exploding? Apparently, it’s nearly impossible for AI.
Alibaba tested 18 AI agents on 100 real codebases over 233-day cycles. They didn't just look for "quick fixes"—they looked for long-term survival.
The results were a bloodbath:
75% of models broke previously working code during maintenance.
Only Claude Opus 4.5/4.6 maintained a >50% zero-regression rate.
Every other model accumulated technical debt that compounded until the codebase collapsed.
We’ve been using "snapshot" benchmarks like HumanEval that only ask "Does it work right now?"
The new SWE-CI benchmark asks: "Does it still work after 8 months of evolution?"
Most AI agents are "Quick-Fix Artists." They write brittle code that passes tests today but becomes a maintenance nightmare tomorrow. They aren't building software; they're building a house of cards.
The narrative just got honest: Most models can write code. Almost none can maintain it.
2
483