Obaid Ahmed

This is amazing. Good engineering, solving problems that impacts globally.

@drawitpoorly me

I wanted to add a whole new feature to a web app and decided to give the task to both Codex and Claude Code at the same time. They are both working on the same task isolated from each other in different folders, db etc. Let's see who wins. 🤣

@Shpigford @browserbase I couldn’t make it work reliably. Ended up switching back to OpenClaw. Interested to see what others are doing to fix this.

okay, what's the secret to getting browser automation to work on hermes that doesn't constantly get tripped up by fancy javascript forms and/or anti-bot tooling? i've tried @browserbase (with residential proxy) to no consistent effect. (hermes is on a mac mini, fwiw)

Stop asking.. "Claude Code or Codex?" The answer is always the same: Both. Picking ONE coding agent in 2026 is like sticking the first search engine you used in 1999 😭 (mine was Altavista 👀) These tools are evolving weekly. The workflows are changing monthly. Nobody actually knows what “best” looks like yet. We’re all just collectively speedrunning the future of agentic software development. The worst thing you can do right now is become emotionally attached to a single tool. Use both. Break both. Ship with both. Stress test them on real problems. Some days Claude Code feels like a literal senior engineer living inside my terminal. Other days Codex suddenly locks in and cooks harder than expected. The people who win this era won’t be "Tool X" loyalists. They'll be the ones adaptable enough to ride the wave while everyone else argues on X

@Shpigford what got you to come around to cold email as sales strategy?

i'm slowly coming around to cold email as a sales strategy 🫣 that being said, WHY IS ALL OF THE SOFTWARE IN THIS SPACE UNIVERSALLY TERRIBLE?!?!?!?! i've tried a dozen options and every. single. one. of. them. is the worst software i've ever used.

google workspace studio is now inside gmail... 7 things every CEO should automate this week (in priority order): 1/ weekly status digest. every report, summarized into a 1-pager every monday at 7am 2/ inbox triage. sort, label, draft replies for the bottom 60% you would never get to 3/ meeting prep brief. last 3 emails, 2 docs, 1 linkedin signal per attendee, 24hrs before 4/ proposal first draft. your template + the discovery call notes, ready before you sit down 5/ pipeline nudge. 48-hour follow-up drafts for every stalled deal in your CRM 6/ expense tagging. receipts in, categorized rows out, every friday 5pm 7/ board update. last 30 days of revenue, hires, pipeline, into the deck format you already use skip the 'AI strategy.' ship 7 working agents in 7 days. the strategy reveals itself.

Most "AI agents" today need babysitting. You prompt → wait → check → fix → prompt again. That's not automation. It's just a different kind of work. I built rundash for the alternative: agents that run on a schedule and only ping you when something needs your hand. Works with 1000+ tools that you already use. check it out at rundash.ai

@ABDULLAHAHMED Must have been a crazy prompt. 🧐

Gemma 4 + @openclaw has been amazing so far.

@wesbos @syntaxfm Working on @tryprovision (provision.ai) to help team build and manage AI workforce.

What are you working on? Send me your project. OSS, Paid, whatever. We're doing a @syntaxfm Syntax Highlight and we will review and/or roast your projects

It’s powerful once you get over the pain of setting it up. I have been working on creating a pain free way to get your own secure, @openclaw or @hermes_agent_ai instance up and running with complete data isolation, its own email address and proper browser. And the core will be open sourced so you can spin up your own personal agent on linode, digital ocean or hertzner or locally via docker.

So Gemma4 with @openclaw has been interesting so far. Anyone else give it a try? All running locally via @ollama on a 16gb Mac mini.

@Shpigford Have you tried the glm-4.7 or glm-5 model?

Okay. But real talk, what model are you using with 🦞? Opus vibes were spot on. GPT has zero character.

@ky__zo @Shpigford That’s what I do. So far it’s working well.

@Shpigford 1) get a Google workspace 2) create one paid inbox 3) add all the domains as aliases 4) create “catch all” or “groups” for specific invoxes Result: 100s of inboxes managed via one inboxes for $8/month

this is a very niche problem but as someone who as an offensive number of domains/projects, dealing with emails to/from those domains is...not great. cloudflare's email routing works great for *receiving* emails to any domain. but replying from those domains is a huge pain.

Everyone's spinning up AI agents.. Nobody's talking about what happens next. You got the agent working on your laptop or Mac mini. Cool. Now the rest of your team wants to either create their own agents or interact with your agent. Suddenly you are stuck. The agent you spun up only works with your data, your processes, your identity. Suddenly you need identity management, email infrastructure, permissions, runtime isolation, and deployment pipelines. That's what we built Provision for. Provision gives your AI agent everything a new employee gets on day one: → Its own identity → Its own email inbox → A secure runtime → Ability for you and your team to teach it specific skills → Native deployment into Slack, Telegram, or Discord Your agent shows up as a team member. Not a bot. Not a webhook. A coworker with boundaries. If you've built something with OpenClaw, Hermes, or any agent framework and you're stuck on "how do I actually deploy this for my team", that's the exact gap we fill. Onboarding a few teams this month @tryprovision Let's chat?

it's so magical to see your @openclaw agent create a full custom application for you .. but to deploy that application so that you can use it outside of your openclaw instance is such a pain. Gotta be a better way. If you are a developer you can probably guide the agent to use render, vercel etc..but what do non-dev folks do?

Blown away with the kind of agents teams are deploying to help accelerate different parts of their businesses. Mostly non-dev teams that are leveraging their ability to have a dedicated skilled agent right in @SlackHQ. We are opening up a few more spots on our waitlist @ provision.ai

@FurqanR @FurqanR I am building provision.ai to solve some of these challenges. Launched mailboxkit.com a few weeks ago as well. Let’s chat.

Agent infra is genuinely underbuilt. Sandboxes, browsers, code execution, auth, file systems, all of it. Almost every layer needs better options. Spent the last few months evaluating what's out there and most of it is very early. If you're building infra for agents I want to try your project.

@0xVikk nope both have their own dedicated workspaces that they work on. Even have their own GitHub accounts.

@obaid 2 devs in one slack, do they ever conflict on the same files?

I am now running 4 OpenClaw agents (2 devs, 1 bdr, 1 seo) to automate parts of my business that typically would take me hours to do myself. What a time to be a builder. None of them are running on Mac mini. Live in my slack. And act like my team members. AI labor shift is happening.