Sabitlenmiş Tweet
David Boskovic
1.5K posts
David Boskovic
@dboskovic
founder/ceo @obvious (prev @flatfile) | schemer, builder, humanist | once described as a “cute guy who likes to play with AI”
Washington, DC Katılım Eylül 2008
1.4K Takip Edilen2K Takipçiler
I swear to god if Bernie comes out with “Claude for president”…
Sen. Bernie Sanders@SenSanders
I spoke to Anthropic’s AI agent Claude about AI collecting massive amounts of personal data and how that information is being used to violate our privacy rights. What an AI agent says about the dangers of AI is shocking and should wake us up.
English
I spoke to Anthropic’s AI agent Claude about AI collecting massive amounts of personal data and how that information is being used to violate our privacy rights.
What an AI agent says about the dangers of AI is shocking and should wake us up.
English
@dboskovic @opencode @cerebras What do you do to review the diff + recording videos ? Also do your design review agents automate critiquing UI?
English
I’ve shipped something like 200k lines of code in the last week and I haven’t looked at an IDE or a CLI
Just specs, QA, and reviews
You are not ready for what’s about to happen
English
we do two layers:
1. purely automated "fast mode" QA on PRs into release branches. This looks like @opencode + @cerebras for fast inference + static analysis + test suites
2. heavy automated + manual review of release branches. claude code agents, design review, security review, agents that review the diff and record videos of them using the feature, etc
English
@dboskovic How are you doing QA? Are you doing that yourself or have you found anything to make that easier
English
David Boskovic retweetledi
If an engineering lead can assign a task to a Jr engineer and then review the PR that person produces then similar affordances should be true for AI engineers.
This gets easier to think about the more removed you are. When AI coding involved an engineer actually writing some of the code this was different
Compliance frameworks haven’t caught up to a world of how things should work when there’s no clear human author but I imagine they will someday. For now, this is a grey area
English
@dboskovic How does this work with regulated industries and similar controls?
If the person controlling the agent can approve their own PRs why do you need approval at all?
English
you probably need to hear this:
- agents should author PRs under an agent username
- any human should be able to review and approve (including human in charge of agent)
this is how you get velocity in agent orchestrated software development pipelines
English
don’t sleep on @tensorlake
this kind of optimization leads to a lot of new possibilities with massively distributed workloads at close to bare metal costs
Diptanu Choudhury@diptanu
github.com/adammiribyan/z… The part about using KVM to get clones of VMs is easy. The challenging part at the moment is cloning sandboxes 100x times across nodes. The bottleneck is around moving bytes across machines. We developed a hybrid approach of moving some data across machines directly, and some through blob stores. Also, requires tuning the network/RPC stack to move as much data the NIC allows you to.
English
any ml engineers and/or design engineers in my little following interested in talking about working on this design tool together? i've got ideas and connections and all that good stuff but building alone is no fun.
x.com/Riyvir/status/…
River Marchand@Riyvir
today’s experiment: chimera. a little tool that let’s you realtime morph through a design system matrix generated by 4 reference images. this one was inspired by listening to @jameygannon talk about her process on Dive Club with @ridd_design and How I AI with @clairevo. her approach to choosing moodboards over prompts made me wonder what might be possible if we applied that same approach to generative UI. and after a few dead ends, the idea turned into chimera. the results are very generic at the moment but I might tune it up if people seem interested. the big reminder for me is how much more inspiring a tool feels when you can explore in realtime instead of waiting for results every time you make a change. lots more things to try in that direction.
English
@dboskovic 😂 I don't think I could have made up quotes this good if I tried!
English
that moment where your execs start posting made up shit in the #general channel to keep your vibe coders hyped 🙄
English
if I only shitpost at 2am will all my new followers
be exclusively in europe?
English
@LionHummer Isn’t this what @cursor_ai cloud agents are supposed to solve?
English
@bleuonbase @obvious It’s not CI. It’s actually shipping the code that’s getting rate limited.
English
@dboskovic @obvious you can swap only ci provider and keep the rest
on gh too
English
so at @obvious we've been developing an autonomous development engine (Autobuild) that is capable of the entire software development lifecycle.
Initially we thought CI would be the rate limiter. Two weeks ago I posted about how we got our CI time down to ~15s.
The rate limit is actually @github preventing us from creating too many PRs per user.
Are we really going to have to find a new git provider that's designed for the future? @olvrgln is this what you're cooking at @mesa_dot_dev ?
@gitlab what about ya'll? Anyone actually ready for the singularity or are we all playin
anybody have suggestions?
English