DevRel & DX Leader ✨ Developer experience is strategy. Educate • Empower • Ship

Portugal
this weekend's touch grass moment: backyard camping for the very first time🏕️
1
1
19
619
there are tiny moments in life when you witness your partner go from dad to hero in your child's eyes. Building the train track to end all train tracks this morning was certainly one of them.
4
1
20
902
some people you meet through work, you’d have picked as friends anyway. 💛
6
1
45
1,056
Lauren Lee retweeted
It is never too late. Reading someone's transition journey, all I could say is just put in the work, the compound effect is the real gain.
2
12
85
2,148
from all of us during our @TBD54566975 days @Square and my time working with @LoLoCoding on ZKP education...this is amazing!!
SEC Commissioner Peirce Calls for Zero-Knowledge Proofs to Replace Data-Heavy KYC/AML Practices SEC Commissioner Hester Peirce said the agency’s Innovation Exemption provides a temporary pathway for tokenized securities to trade through AMMs, helping prevent overseas markets from monopolizing tokenized exposure to U.S. equities while serving as a bridge to longer-term rules. She also criticized the current KYC/AML model for creating ever-larger “data haystacks” with limited effectiveness while turning the financial system into a “panopticon,” and called for zero-knowledge proofs and attribute-based credentials to verify compliance without collecting or repeatedly storing users’ sensitive personal data.
4
4
49
1,630
Lauren Lee retweeted
Design is caring Design is noticing Design is *not* dead Amazing talk by @lmsmonteiro (@pixelmatters_)
4
13
1,233
This is why I'll always read the code and understand, architecturally, how things work vs not knowing and trusting an Agent with whatever it says. Having an understanding of how your systems work, how the systems/platforms you use work, and how it works programmatically is so crucial. Otherwise, you're at the mercy of your Agent. #agenticai #ai #kubernetes
1
2
7
435
This is by far the most fun I’ve had playing around with infrastructure in a while. @flydotio’s Sprites made “just try it” the cheapest option, which is exactly how I want to be experimenting with agents.
Replying to @LoLoCoding
The agent runs on the Sprite, not on my laptop, with the machine as a tool; the shell, toolchain, and broken repo are already there. The whole project, including every mistake, cost about $16. Write-up: dev.to/lolocoding/33-broken-… Tool: github.com/laurenelee/hackju…
1
6
563
lunch while B plays. ☁️ yeah, pretty dreamy✨
11
394
I built a tool to catch confident wrong answers from hackathon submissions. Then I put a coding agent on each broken build to try to fix it, and the agents turned up 6 confident wrong answers in the tool itself. Those "build failures" from last week's post were the environment, not their code. Corrected. Everything ran on @flydotio Sprites, agent included. Here's what the agents fixed, what they tried to get away with, and what else they caught. 🧵
2
1
13
864
The agent runs on the Sprite, not on my laptop, with the machine as a tool; the shell, toolchain, and broken repo are already there. The whole project, including every mistake, cost about $16. Write-up: dev.to/lolocoding/33-broken-… Tool: github.com/laurenelee/hackju…
2
1
577
Part one, where the 33 broken builds came from:
"Does it build?" It's the one question hackathon judges rarely get to answer, because 60 teams in an afternoon leaves us almost no time with each project, let alone time to run it. So I ran all 164 submissions from one event, each on its own disposable computer. 69% of them built. A quarter wouldn't have built on any machine but their authors'.
84
mini roadtrip south to the big city for @lisbonAI_ 🚗 such a throwback traveling to an event just the 3 of us. OG devrel days, back when we'd bring Beau to conferences as a baby. love that we still do this 🤍
1
2
23
763
"Does it build?" It's the one question hackathon judges rarely get to answer, because 60 teams in an afternoon leaves us almost no time with each project, let alone time to run it. So I ran all 164 submissions from one event, each on its own disposable computer. 69% of them built. A quarter wouldn't have built on any machine but their authors'.
2
1
7
668
Ran on @flydotio Sprites. Fresh machine per repo, billed by the second, failures checkpointed so you can open the exact broken state. The whole run was a for loop: all 164 in 55 minutes for $1.01 Write-up: dev.to/lolocoding/164-dispos… Tool: github.com/laurenelee/hackju…
157
nothing makes you nostalgic like restickering your laptop. Each one is a memory, an old team, a conference, a friend I made years ago. ok friends, I spy with my little eye… find yours in here. Which one’s yours? 👇
3
1
18
871
Jev is the AI model that only answers in probabilities. Ran it on my timeline: {"jev": 0.94, "jev_but_skeptical": 0.05, "literally_anything_else": 0.01} confidence: 1.00
3
555