Amplifying human agency @ampcode Prev: founder @sourcegraph
San Francisco, CA
Joined September 2008
- Tweets7.8K
- Following2.2K
- Followers13.5K
- Likes23.4K
You get this for free for any app hosted in an orb btw
I'm now building an "improvements" feature into all of my personal vibecoded applications: a freeform text input that's hooked up to an agent that yolo implements whatever is submitted.
It's an interesting experience -- the activation energy barrier to sanding any rough edge or implementing any new feature becomes extremely low. Someone else who uses one of the applications I created has already implemented 150 improvements over the course of a few days. This pattern feels like a new kind of tool -- liquid software? -- and every application that doesn't have this feature now has a slightly rigid and anachronistic flavor. Why can't I just tell it how I want it to be different?
Now, it's not obvious to me how this pattern can or should scale. Does it become unmaintainable after a while? (Maybe model intelligence improvement outstrips my feature appetite such that it's fine?) How does collaborative software work in this model? Can software designed this way be secure and reliable?
Whatever the larger questions about broader deployment, it definitely works great in a personal context. It reminds me a little of the visions of Lisp/Lisp machines, Smalltalk, and HyperCard -- systems that the user can take control of and modify. Those visions never quite worked, though: the vast majority of software today is not end-user modifiable in any useful way. So, while this "liquid" affordance is a very small touch, the way it changes one's relation to the software feels a bit transgressive and empowering in a Lickliderian and Engelbartian way.
People do this for other processes all the time, in business and in nature. Build a model, see where the theoretical bottlenecks emerge, then see how it holds up in practice. We should see more of this in software.
The trunk-based development debate started by @thorstenball has clearly touched a nerve. So I thought, why don't we look at some numbers?
I built a little simulator with the help of our dear colleague Opus to compare four different integration strategies:
1. Trunk-based development (@thorstenball's horse)
2. Merge queue as the only CI gate (what I'm currently using)
3. PR-based, requiring a fresh base
4. An agentic merge queue, with built-in conflict resolution (dream option, can't buy it anywhere yet)
You can play with the simulator here: lpalmieri.com/simulations/co…
If you're thinking about moving away from localhost dev but want to continue to use compute within your environment, try Amp self-hosted orbs!
> a lot of my opinions on agent performance in general are tainted by months of accumulating & conflicting promptslop
Have been there myself. You really have to clean house and start from first principles every couple months.
If you indexed on the copilot era, you missed the potential of RAG. If you indexed on RAG, you missed agents. If you indexed on TUI, you're missing cloud agents. If you indexed on sandboxes, you're missing orbin' time. Things are moving quite fast.
2 early conclusions from maining @AmpCode for the last couple of days:
- as a harness, it's still at least one weight class above anything else I've tried
- a lot of my opinions on agent performance in general are tainted by months of accumulating & conflicting promptslop: .md files, skills, rules, comments, human docs.
Agressively repairing the opus 4.7-era damage to understandability is a prerequisite for any future incremental improvement
In case you’re thinking of trying Amp coming from another agent
Just realized @AmpCode can import old threads from Codex and Claude Code, so I moved everything over and deleted ~/.codex/ — 43GB freed, and I didn't lose any history.
Amp self-hosted orbs on Oxide compute. Super slick.
Don't run AI agents on your laptop. Don't even run them in a public cloud. Run them on Oxide Computer Company and get the security of self-hosted infrastructure with the scalability of the public cloud.
oxide.computer/blog/self-hos…
The optical cortex can process 10mb of data per second. That's 200x faster than reading text and 1,000x faster than reading a code in a diff.
Time to ask Amp to visualize things.
Charts, timelines, diagrams, mockups: Amp can now build things for you right in the the thread.
ampcode.com/news/visualize-t…
The most cost-effective way of doing iOS and macOS development in the cloud
Amp now has macOS Satellites!
Agents can now spawn Macs and build, test and run your iOS and macOS applications on them.
Rolling out now.
ampcode.com/news/macos-satel…
You have to admire the machinery of capitalism here, orchestrating the labor of the brightest minds and vast quantities of material wealth, all to deliver your next token on the cheap.
And all this is downstream of frontier agents that drive demand for tokenized intelligence.
It was great to trade agentic coding notes with @kentcdodds today and see we share many of the same thoughts about where agents are going.
He onboarded me to Kody and I onboarded him to Amp. Managed to ship a bug fix in our onboarding flow, a small feature request, and put Kody in the MCP registry, all live. Thanks for the feedback, Kent!
Went through this realization over the summer and I can't imagine going back.
When I hear other devs complain about mental overload and exhaustion, I feel a lot is due to context switching between different apps & window management, which essentially go away with orbs.
local development might actually be dead
i've fully reviewed @AmpCode in this video
check it out: youtu.be/TzfaEXJgBBY?si=3CMB…
