Joe Speiser ⚡️
8.3K posts

Joe Speiser ⚡️
@jspeiser
Built & scaled 3 startups to $70m+/yr in ad-tech, ecomm & digital media. Now Co-Founder, CEO @HamptonFounders
NY Bergabung Şubat 2009
382 Mengikuti102.4K Pengikut

@jspeiser The connections just show who you've connected with in the past... they don't show who will respond, or introduce you to the right person.. they don't show relationship. We've been building a new model/dataset that does this. And have an API/MCP in beta.
English

We're hiring a VP of Tech at Hampton.
Looking for an AI transformation leader who's also held together a complex, real-world membership business at the seams.
Both matter. Keep reading.
Hampton runs on a gnarly, battle-tested stack: Stripe subscription logic, Airtable as the db, HubSpot handling the full member lifecycle, all wired together by dozens of automations that trigger each other.
It works. But it's dense. You need to own it, understand every dependency, and make it better without breaking what's live.
AND - separately - you need to lead our AI transformation.
If you've never built real things with Claude Code or Cursor, stop here.
2-3 direct reports. Reports to me directly. Remote US or Latin/South America.
→ joinhampton.com/careers
English

@jspeiser That's good.. but it doesn't have the relationship data... to know the strength of the various connections you have.
English

@jspeiser Many Hampton members are my customers. We can hire for this in central/South America. Holller
English

@JeremyKrak Agreed. That’s when I hand this over to my dev team and go back to Ops and stop pretending I’m technical. 😏
English

@jspeiser I think that level that I'm doing is going to be considered "baseline" in most companies within the next year because the results and output of agent swarming is massive for output of nearly every task (with the right orchestration, and strategically designed management system)
English

@benkeighran PC and Claude are already running those searches for me. Crazy
English

@jspeiser I’ve got a free beta, that searches your relationships (1st and 2nd degree) to try and find the perfect candidate. Lmk if you want to try it for this search?
English

@jspeiser The game changer for me with Openclaw is the full control of persistent memory while being able to utilize all models. Youre on point with the time sink, though. I've burned countless hours wrestling with it.
English

I've been using all three AI agents. OpenClaw, Perplexity Computer, and Claude Cowork.
Here's what I've found so far:
OpenClaw is the most powerful if you're technical, love tinkering and have a ton of time to mess around.
Open source, self-hosted, you pick your models. You can make it do basically anything. But you're also the one building and maintaining all of it (as its always breaking).
Claude Cowork feels the most like having an actual coworker (for 1 specific task). It runs on your Mac, works with your local files, and the plugins for legal, finance, and HR are legit. Downside is it only uses Anthropic models and your laptop has to stay open.
Perplexity Computer is the easiest one to actually get work done (imo) with tons of models running behind the scenes, plugs right into Gmail, Slack, Notion, and it runs in the cloud so it keeps going while you sleep. Tradeoff is you're living in their ecosystem. (oh and their local browser control needs major work, it's unreliable)
so:
Want full control? OpenClaw.
Want a desktop co-pilot? Cowork.
Want to hand off real tasks and walk away? Perplexity Computer.
These tools are changing weekly though, im just trying to keep up.
What did I miss?
English

@jspeiser Well if you spin up a 50 agent swarm inside of one machine, and do it about 4 times... yeah.
I run 150 agents right now on my openclaw system so at that level, its just a different game.
English

@ustechgod not sure yet, I think they will end up being similar though
English

Crazy how evident it is that this was written with AI.
Joe Speiser ⚡️@jspeiser
First week with zero OpenClaw. every routine, every automation runs on Perplexity Computer now. No command line. No babysitting. Time to service has cratered. I didn't switch for the hype. I switched because my time kept disappearing into a black hole. Now it doesn't.
English

@iannwu i think each session is singular, but do dozens of sessions at once.
But im not sure if 1 session was super heavy if it would run multi agents in parallel.
Good q.
English

When I first started in early February, I failed. I waited another week or two, saw some new resources on X, wiped the first attempt, and relaunched.
I am now days away from officially launching a new company that would have been impossible for me to create without Openclaw - or would have cost me $1,000,000 and a year to build. Profound stuff that 95% of people don’t know about/understand. Ppl think AI = chat bot from 2023 still. Silly
English

@JeremyKrak How?! That's crazy. Ive been using it non-stop for 2 weeks and still have 10K left of the 45k.
English

@jspeiser I burned out my perplexity credits in 1 day at $200
English

@jspeiser @WizLikeWizard i'm not going to comment "skill issue" and leave it at that, let me help
first a couple of questions for you both:
- local or hosted online?
- which model?
- which APIs have you equipped it with yet?
- what are you using it for?
English

@spac3man_G After migrating and setting up my routines from Claw my consumption has decreased a lot.
English

@jspeiser How are you affording this, I find perplexity very expensive and very token-hungry. I've got a max subscription and the 10,000 tokens went day one. I've purchased 40,000 more and it was only released a couple of weeks ago. I have to use it very selectively.
English

@jspeiser Your layout of the post offers no concessions. It really feels like a sponsored post.
English






