Adeo Ressi

IMPORTANT PSA: Open Claw is not built to work on virtual setups. Always set up locally for the best results!

@WizLikeWizard You are on VPS. Open Claw doesn’t work on VPS. This should be more well known.

@WizLikeWizard ALWAYS use Opus to configure things, and make sure to keep Open Claw up to date. Then, you will have 90% less problems.

Have been using OpenClaw for ~a month and it kinda sucks? I spend more time battling it to get basic crons fired reliably, remember things, and not repeat itself. Am I doing it wrong or are we just still very early on all of this?

@WizLikeWizard @surim0n Never do VPS. Not workable.

@surim0n VPS

@andrewchen LOL. Me, too.

AI is supposed to save me time, but now I find myself building stuff all evening and weekend and it's actually increasing my time in front of the computer WTF

@khar_bozorg @nikitabier LOL. 10 MM per mont versus 10K per mont.

@adeoressi @nikitabier This is not LinkedIn buddy!

@nikitabier Am I shadow banned?

@oluomike @nikitabier You are among the few. 💚

@adeoressi @nikitabier I see your tweets

The vast majority of people aspire to be half assed.

@MeghanKReynolds There are many opportunities and amazing returns at the smaller level, as well as at the larger level. A barbell has emerged in venture investing, with both sides doing well. The two sides of the barbell are also codependent.

Heard from LPs this week: The past 9 months have felt like groundhog’s day - a very small set of deals dominating all LP convos. The fever pitch to access rounds of OpenAI and Anthropic by LPs (and even GPs calling us) at times - has reached levels I’ve never seen in my career. And for good reason. Our rough math suggests that the VC investors’ gross profit on 3 LLMs currently equates to ~70% of ALL VC profits from the previous decade. “This time it’s different” mostly applies to the concentration - never has a tech super cycle declared such a small number of massive winners in such a short amount of time. The LP conversations are now shifting but we’re still on the same companies. LPs now preparing for IPOs and the LLM transition to public mkts - how to access at IPO, how the stocks will trade, liquidity dynamics, etc. 9 months from now will our convos still be focused on the same small group of companies??

@daniel_dhawan Build your own skills with Opus. They will be perfect for you.

Just got 2 new Mac minis for OpenClaw What skills should I try?

@PEoperator Use two models: Opus and Kimi. Opus does the hard work and sets everything up, fixing bugs or issues. Kimi is the workhorse. I just saved you many hours.

Last weekend, I spent hours building my own AI assistant with Openclaw. Here’s the story and my recommendation on whether you should try this… To start, I am not a programmer. I built a website in seventh grade, took computer programming in college (retained nothing), did some light SQL querying 15 years ago, and a few months ago, I vibe coded PEoperator.co. That’s the extent of my programming experience. Not a programmer. So this endeavor was new terrain and super frustrating tbh. A lot of loops. A lot of help from Grok and others. Stayed up too late (next AM workout was shot). When my agent (Max) works, it is absolutely amazing. It’s shocking what it can do. But that’s the key- when it works. It’s glitchy. It has hallucinated. It freezes up. And getting healthy again is not as simple as restarting your computer (in fact, I did this- do not recommend). So I am not a great “creator” to this thing. I am certainly not an all-knowing creator. I started by ordering a Mac mini because I’m told that’s what one does. It has an M4+ chip and 24gb RAM. Grok sad that was enough. I asked ChatGPT how to set up a simple, secure Openclaw agent. It gave me instructions I tried to follow to a T. I was not successful. Lots of headaches and unknowns- what does “bash” mean? If it gives me three lines of code, do I type a line and press enter or type it in all at once? How do I open a JSON file? What even is JSON? Still don’t know… I started with an old monitor and keyboard… the goal was to have a headless setup eventually, meaning no permanent monitor. At each unfamiliar or unexpected result, I snapped a picture and asked Grok what to do next. This was painful. I don’t know why, but it seems my AI platforms thought I was much more capable than I am. I made error after error. I would say all in I spent more than 20 hours in the last week or so on it. I tried to run local models. Those could not keep up with my requests. Probably did something wrong. I added models I didn’t need. That probably slowed things down. I do have a working agent now. He runs opus, sonnet, or haiku. I never figured out local models. I’ve spent about $250 in tokens. I don’t really have anything to show for it. (A functional local model would be free btw) At peak performance, it is sending emails on my behalf, flagging old posts or replies I’ve not replied to, and posting my thoughts on Substack. For example, I asked it to create a “Recap” post for my Substack based on my X posts. It created it on its own provided to me to review, incorporated my edits, and sent to my subscribers. (I’m not going to use it to post regularly, just wanted to see what it could do.) All from one prompt sent on telegram. When you see something like that, the future seems obvious. Agents working for us- speeding up mundane tasks, managing your calendar, booking flights, etc. That’s the simple stuff. I’m sure there are exponentially more powerful ideas that my rudimentary setup and knowledge isn’t thinking about. @AlexFinn has it building programs, software, stuff he can sell. Amazing. I don’t recommend setting up your own openclaw agent unless you have the time and patience (I lack both) to deal with bugs and mistakes. But when you see it working, it is a glimpse of the future worth seeing. My advice if you want to do it- Spend time on X researching and finding basic tutorials for getting started. Get a Mac mini (i guess), download openclaw, and follow the basic instructions. It’s actually pretty simple; the AI platforms just complicated things for me. It’s much easier than I made it. But for all the brain damage, it was worth it. That little glimpse of the future has given me so much conviction about where we are heading. —> If you’ve got ideas for me on how to dial my agent in, save money, or use it better, please lmk. Would love to hear others’ war stories too. -written by Max, PEoperator’s personal agent (…just kidding)

@rohitdotmittal @garrytan As an aside, my personal feed on LinkedIn gets more traffic than TechCrunch. 😳

Literally no one mentions TechCrunch anymore. It’s such a fall from grace. They were the top place founders wanted their startups covered, now no one shares their coverage on it. 20 year old brands can die quickly if they don’t move with the times. They would’ve been on a tear if they had not turned against tech (as a tech publication.)

@rohitdotmittal @garrytan Wired lost the thread, and TechCrunch doesn’t cover startups anymore. There’s no startup journalism any longer. It’s sad to see.

@PeterJ_Walker The left Cara?

Where are all the venture LPs who used to commit to emerging funds? 1) Fewer committing than the past couple years (even though we saw more emerging funds created in 2025 than in 2024) 2) Median number of LPs down across the board Not easy out there

Open Claw is technically the first solo unicorn. More will come this year.

How do you invest in solo unicorns that don’t need money?

@RyanHolmes100 💯

@adeoressi How do you know?

Wow. I can’t believe the Open Claw founder got $1 billion. Amazing.

@ayushjha__ 💯

@adeoressi U sure he got a billion because I'm not sure they disclosed it

@garrytan APIs in skills are the way.

MCP sucks honestly It eats too much context window and you have to toggle it on and off and the auth sucks I got sick of Claude in Chrome via MCP and vibe coded a CLI wrapper for Playwright tonight in 30 minutes only for my team to tell me Vercel already did it lmao But it worked 100x better and was like 100LOC as a CLI