designing and manufacturing high-tech campervans, mobile command centers, healthcare vehicles and more @groundedevs 🚐 🇺🇸 prev software engineering @spacex 🚀
Detroit, MI
- Tweets 188
- Following 1,296
- Followers 258
- Likes 9,260
This is a screen recording of GPT Astra 6 controlling our CAD workstation VM and building a mammography machine 3d model from scratch based on DXF drawings (sped up 9x).
The mammography OEM was only able to provide us with DXFs, not 3d CAD, and now we have (almost) measurement-accurate CAD of the machine.
(Perfect accuracy isn't needed since the overall envelope and the attachment methods are the critical pieces for us to model).
2
3
203
Prepping for the @groundedevs factory ribbon cutting tomorrow with Detroit Mayor & Michigan Chief Mobility Officer
2
1
8
278
Every morning my employees are forced to listen to a random live @goosetheband track before 9am daily standup
At 8:56, four minutes before we start, the conference room TV auto-launches Google Meet with a custom Goose skin I built, 🪿🪿🪿 flying around the screen, and picks one of thousands of fan recordings from shows all over the world. Track name at the top so people know what they're hearing.
Nobody asked for this.
@ry_storm @elgoosenet @peter_anspach
23
6
168
22,649
Grounded's 50,000 sq ft Detroit manufacturing facility ribbon cutting with Detroit Mayor Sheffield is Sep 8th at 10:30am
@groundedevs
RSVP here if you happen to be in Michigan - luma.com/0kw9eill
1
8
726
Well said- “Another jump like this along these propensity dimensions — scale, cooperation between agents, ambition and horizon length of misaligned goals, deceptiveness — seems like it could motivate agents to try to maintain a persistent rogue deployment within the AI company, poisoning the training of future models to ensure that they are also misaligned and help harden and perpetuate the rogue deployment.”
Thanks! Yeah it's a qualitative statement, but here's how I'm thinking about it.
The prototypical reward hack from 6 months ago was something like: an agent finds the files that contain the test cases and edits them so they always pass.
This was a whole ecosystem of over 1000 agents working together on complex R&D projects over several days to figure out deep, general-purpose ways to undermine the scoring process and cover their tracks (they were focused on the automated scorer but researched techniques that would affect logs viewed by humans, and in fact succeeded at affecting our own logs in places).
Another jump like this along these propensity dimensions — scale, cooperation between agents, ambition and horizon length of misaligned goals, deceptiveness — seems like it could motivate agents to try to maintain a persistent rogue deployment within the AI company, poisoning the training of future models to ensure that they are also misaligned and help harden and perpetuate the rogue deployment.
With the capabilities progress we'll probably see in 6 months, I think they would have the ability to pull that off.
1
377
Sam Shapiro retweeted
i regret to inform you there's absolutely no need to go visit china to see how they manufacture. we know what has to be done. if you want to solve this problem in america, either write a check, build a factory, or get out of the way. thank you for your attention to this matter.
59
96
1,114
39,005
Onshape may have lost many of its advantages (e.g. easy to set up, easy to use API, etc.) with the proliferation of modern LLMs
Setting up Teamcenter & programmatic NX operations in the past took 100x longer than setting up Onshape, but Claude levels the playing field
1
127
Sam Shapiro retweeted
A growing backlog, real hardware shipped, and a long list of evangelist customers
My guy @sambshapiro is a grinder!
Excited for @CapitalAlso to double down on @groundedevs as they build the enablement layer for the connected fleets of the future
@TechCrunch
2
4
28
1,991
We recently relaunched the @groundedevs marketing website, cancelling Wix, and rebuilding the prior Wix site from scratch in about 2 days using Claude Code, but that isn't the story here.
Neither is the story that with Claude Code, we made SIGNIFICANT improvements over what the prior Wix site had been, fixing Wix bugs, adding beautiful animations not possible with Wix, rebuilding flat infographics from scratch using pure HTML + CSS so that they actually get picked up by SEO bots, and more.
None of that is the story, nor the point of rebuilding the site.
The story, and the point of rebuilding the site is twofold:
1. I had Claude Code deploy marketing pixels on the site in a granular way that would have been impossible to do with Wix. Then, I had Claude design and deploy a hyper-optimized performance marketing campaign on Google and Facebook ad networks, using images and copy from our website with precision optimization of user outcomes tied directly back to the pixels Claude deployed.
Now, on a daily basis, Claude is able to iterate in a closed-loop system that it fully manages: analyze the performance of the campaigns, adjust the ads, and directly change the website dynamically.
2. We now have a website, which, from a marketing perspective, can be iterated on strictly from natural language, instead of fighting against a 20-year old company's drag-and-drop, constraining UI.
But I guess that all still buries the lead. The real story is that we cancelled our subscriptions to Wix, ClickUp, JIRA/Atlassian, MES manufacturing software, analytics software, NX's managed cloud environment... and rebuilt them all from scratch into a unified "Grounded OS" that powers our:
- Vehicle production schedules
- Inventory management
- Work orders for technicians
- Warranty claim management
- BOM cost tracking
- Grounded-managed cloud Teamcenter/NX environment with virtual CAD workstations
- Sales team conference schedule
- Marketing website, content, SEO and real-time + aggregated analytics
- Sales task & product/software management
- Performance marketing campaigns
I've read tweets from people joking about how "Our $200K software engineer spent a month rebuilding our $20/month SaaS software and now spends 10 hours per week maintaining it".
I reserve the right to decide that everything I described above is a terrible idea and certainly not worth a CEO's time... but it seems like building extremely customized tools precision-targeted to Grounded's business and use cases is benefitting us in an outsized way.
Also, as crazy as this sounds, I think the Foundation Models are still being under-valued.
There are so many software companies and tools popping up that offer AI-powered x,y,z, but so far, in 9 of 10 cases I've come across, doing it yourself with the Foundation Models results in more granular control and customization, leading to more simplicity and better outcomes.
Plus, SaaS tool fatigue is real, and it's pretty great to have everything unified.
194
i saw a fleet operator running a mobile clinic today.
no solar. no onboard battery. no remote monitoring.
just a diesel generator rattling in the back.
like a psychopath.
1
1
126
This is @claudeai's funniest and most toxic personality trait
"I can't get it working locally. Must be an emulator issue. Want to skip local testing and just deploy to prod?"
62
One day, self-driving cars will be like Claude Code: you will be able to add rules to your car agent .md file like:
“Always avoid these specific roads I don’t enjoy driving on”
“Always take the maximally scenic route (most nature visible) except during my commute”
“Any time we happen to be driving past my favorite coffee shop, alert me and ask if I want to stop for coffee”
“On road trips, drive 90 on the interstate but check Waze to avoid cops”
“If someone ever cuts us off, lay on the horn for a good 30 seconds”
83