
Founding Engineering at 7AI | Previously at @stripe, @vmw_carbonblack, @ravelinhq, and @RSAsecurity. He/Him. Views are my own.
Boston, MA
Joined May 2008
- Tweets2.3K
- Following893
- Followers731
- Likes27.5K
Ereli retweeted
Let's play a game - win32 types vs Polish language:
LPCWSTR
PSZCZYNA
WCSLEN
WCZESNY
LPCTSTR
BYDGOSZCZ
WSTRZAS
HGDIOBJ
DOWOD
HWINSTA
DLUGOSC
LPCSTR
DWORD
KAL
LPWSTR
SZCZECIN
PCWSTR
BLAD
PUHALF
CHUJ
UHALF
Ereli retweeted
When I politely decline scheduling a "quick call", it's not because I don't literally have the time — there's always room for 15 minutes here or half an hour there! — it's that I can't afford to spare the attention.
Ereli retweeted
So much corporate red tape can be traced to a single anecdote where someone fucked up and their manager didn't have the balls to tell them directly. So they made a policy instead. Don't be that coward. Just tell Ben.
Ereli retweeted
Turso is an incredible technical feat. A Rust rewrite of sqlite, with an async-first architecture, incoming support for concurrent writes, vector search, and browser / wasm support out of the box.
I think this has a very good chance of being a foundational piece of infrastructure of the vibe-coding age. On-demand, sqlite-compatible global databases that can also run in-browser and on-device.
The pace at which the project is evolving is most definitely *not normal*. @penberg and @glcst are built different.
Demo: shell.turso.tech
Ereli retweeted
If recruiters said you were 'too much of a generalist' or 'hard to put in a box' — your time is now. As AI commoditizes specialized tech skills, liberal arts thinkers will drive the next wave of innovation. We're trained in the Socratic method, synthesis, and rapid learning. Give us 48 hours on any topic and we'll absorb enough to engage meaningful conversations with true experts and now AI agents. Specialization is for insects; adaptability (AQ) is for humans in the age of AI.
Ereli retweeted
Why can AIs code for 1h but not 10h?
A simple explanation: if there's a 10% chance of error per 10min step (say), the success rate is:
1h: 53%
4h: 8%
10h: 0.002%
@tobyordoxford has tested this 'constant error rate' theory and shown it's a good fit for the data
chance of success declines exponentially
Ereli retweeted
Two needles in a haystack: Our latest study explores how LLMs perform on the same task with different lengths of context.
Accuracy dips when models must not only find, but reason over two text parts. Even on just 3000 tokens!
Results and analysis 👇(1/7)
arxiv.org/abs/2402.14848
Ereli retweeted
Why did I copy and paste my password from 1Password into a website without letting the browser extension automatically populate it? Because time and time again - like right now - orgs use totally different domains that don't match the one you registered on.
Ereli retweeted
If the Trump administration really wants to compete with China… make this a rail yard in the USA 🇺🇸
Ereli retweeted
"Claude, create the worlds most annoying CAPTCHA"
"Make it more annoying"
"Make it REALLY annoying"
(apparently this is all technically solvable)

































