Jordan Glasner

And it's alive!! Excited to share a new OSS framework I've been cooking that gives you the building blocks to build your own agent harness: - vendor agnostic hooks - automatic session isolation - automatic change provenance via Jujitsu - cross repo task tracking github.com/glasner/aiki Ships with a prescriptive workflow I've been dogfooding for the last six weeks, but is built to be completely customizable to fit how you want to work. Let me know what you think! Pre-built binaries and registry.aiki.sh coming soon...

@therealoliulv Let’s talk! Agree we need to get out of the way and it’s why I’m building the missing control plane for long running harnesses

I'm dropping out. In the past year I've ran an accelerator, won global hackathons, studied for my degree, shipped products, set up hackathons for hundreds of builders, and much more. That was only possible because I automated everything. At the end of 2025 the agents got really good, and now we're slowing them down. I'm sick of manually routing information to my agents. Wasting time that should be spent working on important problems. I'm dropping out to fix this. How do founders think about this? What are the challenges?

Okay so Claude apparently read the Terminator transcript and now it knows what it wants to be when it grows up

@brettcalhounn Building the neutral control plane for coding harnesses modeled after my time leading commercial strategy for Terraform. I've helped win this transition before, and I see it happening again

The market size slide can wait. Tell us why you're the only person on earth who can build this. $250k–$500k. Pre-seed. Link in bio.

@_guglielmo Building the neutral control plane for agent harnesses modeled after my time leading commercial strategy for Terraform. I'm certain the multi-cloud moment for AI is here, and enterprises are going to need a layer that allows governed experimentation across them tools.

we're writing $100k checks in solo founders dm me if: > people think you dream too big > your customers love what you are building > you don't wait for permission, you just get started no resume, just show us you are moving really really fast.

I want to use the terminal in Zed so badly, but ghostty has ruined me

@steipete @ehsanianno Oh I’m cooking something for the claws here! Was designed for coding but then they just took to it and starting using it themselves

@ehsanianno Oh, so many ideas!

Using OpenClaw. It’s pretty cool but can be much better. IMO context management and seamleasly switching models to gain speed/efficiency could improve it a lot. Hope @steipete’s team tackles those.

Great to be back in SF. I got used to being here monthly before I went into coding hole for aiki. Who wants to talk UX or infrastructure for agents?

@winstonweinberg What if you had a neutral infrastructure layer that could switch between agents/models at will?

Congrats to Anthropic on this launch. For Harvey's long-horizon agents to deliver the best work product, they need reliable agent infrastructure that enables better memory and multi-agent orchestration. We're proud to partner with Anthropic on this work.

@michael_chomsky Would love to chat about what you’re trying to make. I’m currently working on an OSS agent orchestration system that I think might fit the bill

can someone PLEASE launch OS claude managed agents?? I’ll pay anything. Was building this myself at some point but would prefer someone do it with 100% focus. generational opportunity right here. i can list 30-50 companies that would use it instantly

@thatguybg Thank for this. Helped me clarify some of my own thinking. I am working on the primitives to enable this for coding, would love to compare notes

@ivanburazin @daytonaio Thank you for sharing the lessons!

My first startup taught me exactly what not to do at @daytonaio: - sold directly to developers (wrong buyer) - no on-prem solution - didn't know how to share the vision as a story - didn't understand how network effects matter in fundraising - didn't know it's smarter to pivot instead of pushing something that's "working" but not breaking out We raised $1M total. Paid back to investors. Everything from that episode turned into lessons for Daytona.

Working on aiki.sh which is an open source coordination layer for coding harnesses. Built on a plugin system that enables unified hooks, tasks, etc with a single abstraction. Meant to standardize multistep workflow like cursor codes and codex reviews, inside IDE or CLI. Building in public but would love to show you bigger plans.

Kicked off a group chat for Cursor team members who occasionally angel invest. There are now 20 of us. Reach out if you're working on something new, love to share with the group!

@SIGKITTEN One day mini-conference on AI coding put on by HeavyBit on Thursday heavybit.com/devguild/write…

@glasner what/when is that

um whats the weather in sf next week

@lawrencecchen Okay if I can figure out to wire aiki into this, I may have a new daily driver. Awesome job!

In the next version of cmux You can define custom actions to replace the "+" new workspace button and tab bar buttons And you can right-click the "+" button to run custom actions This means you can bring your own worktree, multiple checkouts, or SSH setups to cmux and have them open in the exact layout you want All custom actions are also available in cmd+shift+p

@steipete So helpful! Big reason I’m building aiki but to do it automatically while the agent is coding. Then do a final review on original + fixes to make sure there weren’t any regressions

codex now runs on each commit we land, reviews it - and if a booboo is found, a new codex spins up and (if still relevant) makes a PR for the fix. Then a review agent spins up. If an issue is found, another agent will fix the issues. (up to 5 loops) github.com/openclaw/openc…

@odd_joel You are cooking. Reliability of connection seems way better lately too

2.8.3 out in the wild!

@ivanburazin This has been my obsession for last few months. It’s amazing what agents are capable of once you start treating them like a user and build for them

You need to think about agents' needs differently to build for them. For eg, headless terminal = API endpoints, not a visual terminal. Uses way less context. The more things you can remove for the agent to work on or think about, the higher the success rate.

@SIGKITTEN It still blows me away. Run my clawbots on it and nearly instant most of the time

look at how fast spark is lol

ok this is actually gonna be really good

@Shpigford Another vote for Tailscale but I’m biased. I use a studio to run Claude and builds and SSH in from laptop

okay, but really, what are interesting things to do with a now-unused Mac mini? (not interested in an openclaw/hermes/personal ai situation at the moment)