here for digital technology stuff
OASIS
Joined May 2021
- Tweets65.7K
- Following6.1K
- Followers4.4K
- Likes23K
foxriverdevops retweeted
agents are fundamentally non-deterministic by nature. you cannot prompt that away. so any agent you give open-ended spending power to will eventually do something you did not intend
non-determinism plus open authority is unbounded downside. you do not fix unbounded downside with better intentions. you fix it by making the downside bounded by design
that is the entire product thesis of VAA
foxriverdevops retweeted
I wrote down how my work with agents changed this year. I still break down the work into vertical slices that ship on their own. I rarely read diffs now: independent reviews and a risk check catch most of it, so my time goes into shaping and guardrails.
mgratzer.com/posts/reforging…
ALT Two workflow diagrams side by side. February, a human in the middle: I plan the issue with the agent, the agent implements and self-reviews, I review the PR with line comments and the agent fixes it, then I merge. October, a human at both ends: I shape the issue with the agent. Inside guardrails from the repo and Forge, the agent ships, guards and lands the PR with a risk verdict. Low-risk PRs merge automatically, the rest come to me with a list of where to look. I shape the system, turning drift into new guardrails.
foxriverdevops retweeted
you wouldn't hire people you can't trust. so why are we giving root access to agents we can't audit
foxriverdevops retweeted
AGENT SWARMS ARE HERE
*To accelerate your business
Today we're launching Agent Swarms @grepdotai to accelerate your business by tackling the backlogs your team can't keep up with.
Swarms are teams of agents that work together to automate a task queue in parallel.
foxriverdevops retweeted
good time for this oldie: youtube.com/watch?v=u2DLlNQi…
foxriverdevops retweeted
so many gems and great tips in this talk from @yminsky about how to improve your codebase for agents
• agents thrive on feedback, and types are one of the best kinds
• make illegal states unrepresentable with types, and it enforces global rules across your codebase
• property based tests, fuzzing, and even full-system deterministic simulation testing
• expect tests: just print stuff out and capture it in the test. makes behavior legible. it's very similar to how compilers have fixture tests, it reduces the review surface area
• formal methods used to be very expensive: seL4 is about 8,700 lines of C and took around 25 person-years, with 23 lines of proof for every line of C code, but agents change the roi
youtube.com/watch?v=rUYP4C29…
foxriverdevops retweeted
Agents reinvented how we produce code. It changed our tools and half of our workflow. It’s time for the other half to catch up.
Today, we’re shipping Entire: the Git forge for agent-human collaboration. We’ve built @GoEntire using Entire. Now we want to see what you'll build with it.
Here’s why and what’s different. 🧵
foxriverdevops retweeted
Entire is the ONLY company I've invested in the last two years: b/c it's clear that we need an "AI-native" GitHub, and also clear that GitHub cannot disrupt itself.
Entire just reinvented the PR for agents with Trails, and offer a Git forge that can handle agent load:
Agents reinvented how we produce code. It changed our tools and half of our workflow. It’s time for the other half to catch up.
Today, we’re shipping Entire: the Git forge for agent-human collaboration. We’ve built @GoEntire using Entire. Now we want to see what you'll build with it.
Here’s why and what’s different. 🧵
foxriverdevops retweeted
#Sui will become the place where agents coordinate and transact at scale
VAA is just the beginning
foxriverdevops retweeted
The best vid explanation of LoRA I've seen. Amazing work @juliarturc 🫶
Over the past few days, I dug into the geometry of LoRA to understand how exactly it constraints the geometry of the search space. There's a beautiful connection to the *intrinsic dimension* of a task.
Also learnt I'm saying "difficulty" wrong... 🤡
0:00 MNIST vs the walking humanoid
1:05 Do we still need fine-tuning?
3:05 What is LoRA?
6:55 What does the rank constraint actually do? (geometric intuition)
9:50 Intrinsic dimension: training along random directions
13:30 Intrinsic dimension of pre-trained LLMs
15:10 Demo: fine-tuning a query router with Qwen3.8-27B
foxriverdevops retweeted
software engineers are going to have to tackle a new set of problems to unlock powerful multiagent systems to solve the world's hardest problems - right now is one of the most interesting moments in history to be an engineer.
foxriverdevops retweeted
"but this is only the beginning. we are collaborating on much much more"
i believe people all over the world will use #Sui to move money every day. that's what we've been building for from the start. and that's what makes the partnership with samsung such a huge milestone
but this is only the beginning. we are collaborating on much much more
super pumped about what we can do from here
foxriverdevops retweeted
"One of the biggest stories from Sui Basecamp in Singapore: Verifiable Agent Arbiter (VAA), an evidence layer for AI agents being developed by @Mysten_Labs in collaboration with Google Cloud."
VAA also links with Sui Agent Payments, so when an agent spends money, the authorization and spend decision are on the record too.
Initial enterprise deployments are planned ahead of broader availability.
Full announcement: mystenlabs.com/blog/mysten-l…
foxriverdevops retweeted
this is a new chapter for the Sui Stack! LFG 🚀
1/ Sharing the first update in a new quarterly series from @Mysten_Labs – take a look mystenlabs.com/blog/letters-…. In it, I lay out my thoughts on:
What problems need to be solved in order for agents to be trusted by humans and to trust each other.
Why I’m convinced Sui and Walrus will be fundamental in solving those problems, and why Mysten Labs is the team to tackle them.
2/ We’re also setting up a new, dedicated AI business unit within Mysten Labs, which I will personally lead, in addition to my responsibilities as CTO. At Sui Basecamp, I gave a glimpse of what we’re working on: youtube.com/watch?v=rCWnarkg…
3/ To ensure that the team has full time and bandwidth to dedicate to our agentic strategy as well as supporting the Sui and Walrus ecosystem, I have asked @KevinBoon_ML to step into the CEO role. Kevin has been instrumental in leading Mysten Labs alongside the co-founders from Day 1, first as our Chief Legal Officer and then as our President. I planned for this day when I made him President, and have the utmost faith in him as my successor and our new CEO.
Now, back to work.
foxriverdevops retweeted
Brutal truth from @Lethain
“If you're in [middle] management (…) you sort of have to decide, am I retiring now or am I re-skilling?”
He did just this: step down from leading a large team (250 ppl) to be hands-on with a small team (25 ppl), ship code & re-learn the new “rules”
Why a 3x CTO went back to a 25-person team. From Will Larson @lethain, Imprint CTO and 4x book author.
"It's a hard time in middle management, right? There's a lot more competition for these roles. What I found personally very rewarding is getting back into the details of writing software.
When I left Carta, my team was around 250. When I joined Imprint, the team I was leading was 25.
This a unique moment to actually re-skill. If you're in management, I think right now either you have to get back into the details, or the world is changing so radically that the profession you know will not be valued in three years. You sort of have to decide, am I retiring now or am I re-skilling?
I think there has to be some openness to taking a backward step or a lateral step. What I knew I wanted was getting back into the details and relearning how to do software in the current age. That's what I believe is going to be the unlock to the longevity of my career."
