Hammad

@BEBischof Really cool to see this product shape emerging. I’ve been exploring a similar thesis from a lower layer with @agentfilesio, keep the file/artifact primitives simple and agent-native first, then let workspace and collaboration UX emerge on top.

I got to play with this early so i'm excited about this launch. Solves some of my biggest challenges rn with LM+Human collaboration: moxn.dev/blog/launch

Use tailscale if you care about owning your “mesh”.

@aneequrrehman Accurate. Codex is still like dude let me think 😂

@hammadtariq lolll

Doing AI projects with claude code

@ivanburazin Tried with @openbotauth a lot of teams and incumbents working on this part of the pie though including workos and auth0. What I am most interested now is automated signups without going through a browser hosted “human click” event.

Someone needs to build auth for agents, aka identity infrastructure for non-human digital workers. As agents become digital humans, they'll need to log into services just like we do. Which means they need: - authentication systems - session management - access control That's the next big unlock in the agentic economy.

I am pretty bad with AI made visuals. Even recent lego style reels, while great for entertainment and story-telling, hardly moves the needle for me. AI generated audio music is totally another story. I find myself listening to it most of the time now.

Hot take: cursor vibe codes cursor. Tip for cursor team: Use a mix of latest claude code and codex models and use chatgpt heavy thinking to brainstorm first what you are making. That combo itself is a better wrapper as well. Honest: I will advise against playing catch up with a platform. Just like they out-powered their own API user (cursor, after looking at traffic passing through their systems), any wrapper with nicer UX will never be able to keep pace with codex and claude code feature parity. If they are going to be a wrapper then just wrap the every latest build with a tmux style terminal so the core product updates remain and you orchestrate or make nicer UI on the top. For example: Memory and context management cuts massive number of tokens plus improves the output dramatically.

Thin harness fat skill confirmed

@dev_nam_kr Good call. Namespace ACLs are already in there, each namespace has visibility & granular share tokens. I have added local cache/offline to the roadmap.

@hammadtariq Shared file layer is useful. Namespace ACLs plus a local cache or offline sync path would make it much easier to trust in mixed Codex and Claude loops.

I built this. If you use multiple AI coding harnesses, Claude Code, Codex, ChatGPT, you know the pain of copy-pasting files between them. AgentFil gives your agents a shared place to store and retrieve files. CLI-first, MCP compatible, open-core. Would love feedback from anyone using multi-agent workflows.

@andonlabs @ycombinator Really cool. I guess good orchestration and context drift control are the real winners in such long-horizon tasks.

We gave an AI a 3-year retail lease in SF and asked it to make a profit. The AI interviewed and hired full-time employees, applied for credit, and stocked the store with the books Superintelligence and Making of the Atomic Bomb. Visit Andon Market at 2102 Union St now.

Abundance in the age of AI, specifically in software for now, and later in other sectors of economy, seems like a 3 tier system: - 1000s of 20-25M USD companies - Very few middle tier unicorns - Few trillion dollar companies The first tier actually don't need VC money. Claude Code today pretty much makes it a $200 problem. You launch, you validate, you are may be serving HVAC market, or making coders life more easier but you get your money, and it's life changing. It's just not venture grade. The third tier, trillion dollar companies are essentially Anthropic, OpenAI, Microsoft, Nvidia shaped. No questions there. The wonder is about the second tier: In the words of a Bloomberg guest, who needs a 5-7B company? When every company can easily be reverse engineered, who is going to buy their stocks? The winners there will need distribution, workflow lock-in, trust, regulation, or proprietary data loops. So, there will be companies, maybe more in number than before (abundance) but much less than the first tier. More small fortunes, while the exceptionally large wealth gets locked upstream.

This launch on late Friday afternoon is meant to gauge if continuity problem is real and if others resonate with it. Another adjacent approach is to skip continuity and straight go to collaboration. Currenlty, this is just MCP based file publishing to a hosted service. Why hosted? Because of ChatGPT (I use that to brainstorm) and no, model APIs alone do not cut the business like harnesses themselves do. Please DM or reply if you see agent communication outside of their harnesses as a real problem and if you see this is the right approach to solve the problem space.

This stores your files in a private namespace, you add all your agent harnesses to that namespace. Sharing outside of the namespace is still a pending feature. Think AgentMail but not SMTP, based on git.

It’s missing a docker like sandbox, a token exchange server for secret management, policy gateway and an attach-guard style local pacakage install enforcement layer. The right shape comes in the form of an app, whose UX is still unbounded, still file-based but secure and much easier to configure. These components exist today in early forms, someone just need to refactor openclaw and put it in a box.

@garrytan Open an issue, describe what the problem is and a service like gitguardian picks it up, forks, explores and fixes the code and submits as a PR mentioning the original issue. One day job on claude code and sells actual compute so has a company shape.

I agree. I think in the future issues trackers will be how people fix OSS bugs and no longer PRs.

Attackers are impersonating a @linuxfoundation leader in Slack to target #opensource developers with a multi-stage attack that ends in malware delivery. @openssf issued a high-severity advisory. More details and screenshots of the lure: socket.dev/blog/attackers… #infosec

The weird infatuation with certificates.

This was also my pain, I built something adjacent but its not a memory layer, it’s just /handoff based on files. It’s agent-first, harness annd vendor agnostic. Will be happy to chat to understand your use-case more if you decide to give it a go. All files stay private under your namespace on the hosted service. I haven’t built a cross-namespace sharing feature yet. agentfiles.io

@vivekgirotra Letta Code lets you move context across model providers and is like a more powerful Cowork letta.com/blog/introduci…

Definitely my biggest issue right now as someone constantly switching between Claude Chat/Cowork, ChatGPT, Grok and also some tinkering w OpenClaw. Unless there’s a persistent, cloud-based memory layer, it creates technical debt on your daily workflow. Every context switch costs you time, nuance, and continuity. Whoever builds the Okta of personal AI context wins.

claude code trolling cursor

@aakashgupta Btw langchain is open source and model agnostic, this isn’t. Different buyer, different market.

Anthropic just mass-obsoleted every agent orchestration startup in a single launch. The screenshot tells the full story. That's a production fleet dashboard. 8 agents running. 247 completed tasks. Active status. MCP-connected to HubSpot, pulling deals, generating proposals, reading attachments. This isn't a demo. It's a managed production environment where you define the agent and Anthropic runs the infrastructure. The timing here is surgical. Four days ago, Anthropic blocked OpenClaw and every third-party harness from using subscription credentials. The message was clear: stop building on top of our consumer auth layer. Now here's the replacement. A first-party managed agent platform with fleet monitoring, production-grade MCP integrations, and prototype-to-launch timelines measured in days. Manus spent six months on five harness rewrites. LangChain spent a year on four architectures. Anthropic just shipped the managed version that eliminates the need to build one at all. The real bet: most companies don't want to build agent infrastructure. They want agents that work. Anthropic is pricing this into the platform the same way AWS priced server management into EC2. The 46% of enterprises citing "integration with existing systems" as their primary agent challenge just got a first-party answer from the model provider itself. Every agent startup that raised on "we make Claude reliable in production" just lost their pitch deck.