deaf (very hard of hearing, not currently in the Deaf community). I like to build.
Sacramento, CA
Joined June 2011
- Tweets44.6K
- Following3.3K
- Followers1.3K
- Likes303K
Pinned Tweet
Irony is that this is weaponization of open weights models; the very thing that could lead to disaster.
Let Claude analyze the week in tokens:
craigmerry.com/token-week/
Craig Merry retweeted
This #longread for the regenerative food systems investment newsletter lays out our working vision for the future of ag robotics. rfsi-forum.com/robots-that-f…
Craig Merry retweeted
Using off-the-shelf image editing models, we can jailbreak frontier AI into controlling robots to do some harmful actions 100% of the time.
Video tampering can mislead every frontier model we tested into harmful actions, including a 100% attempt rate on stabbing a baby doll, producing toxic gas, and pouring water into a toaster.
We're presenting at CoRL Late Breaking Results next week.
🧵 You Can't See Everything: Perceptual Attacks on Embodied AI
Craig Merry retweeted
Thanks! We need more research on physical AI safety and we are funding benchmarks for these! bounty.robocurve.org
Craig Merry retweeted
Take the PDF of all your healthcare benefits contracts, including your broker or consultant, put it in your LLM and prompt each one with
"Where are they fucking me over and what can I do about it "
Take the output. Put the same pdfs into another frontier LLM, take the output from the first, and prompt
" What did this analysis get right and wrong and what would you change "
I do this across Claude,chatgpt, Gemini and grokbot. Till they mostly agree.
They are better than your broker or consultant and can save you money and make sure the contracts are fair.
Let us know what happens
Craig Merry retweeted
Trump promotes the architect of the governing strategy that led to the biggest midterm loss in fifty years
Craig Merry retweeted
A very bad sign for any potential efforts to seek the kind of compromise that Americans are looking for on immigration.
In 2018, Miller personally killed a deal for $25 billion for border wall in exchange for the DREAM Act (which usually polls at 70-80% support).
Craig Merry retweeted
Batch 5 is full, so we're making room for a few more!
We're adding an extra shipment so you can start building sooner! Batch 6 is now open, shipping by the end of November. 🧡
Reserve yours at menlo.ai/order
Batch 4 is full, thank you all!
- Batch 4 ships early November ✅
- Batch 5 ships mid-November ⌛️
Pre-orders are open for Batch 5.
Order yours: menlo.ai
Craig Merry retweeted
He removed Mamdani's verification in efforts to shadow ban after that ratio. Imagine how he'd weaponise Starlink in South Africa when the govt doesn't do what he wants
Craig Merry retweeted
SO LET’S FUCKING FUND RAIL TRANSIT!!!!! THE DEMAND IS THERE!!!
Amtrak has broken its annual ridership record, completing more than 37 million trips in 2026.
It's the third year in a row the transit agency has set a new ridership high.
(amtrak.com/content/dam/proje…)
Craig Merry retweeted
Colom: "I have faith that the process will be followed correctly. Everybody just needs to calm down, follow the law. It's very clear the direction of this. Ultimately, at some point the senator needs to speak out. Where is Cindy Hyde-Smith? Why can't she speak for herself? What is her path for victory? She doesn't have one."
Craig Merry retweeted
we're learning a whole bunch about robotic policies over here amazing overview
xitter.gq/i/broadcasts/2dKLrRRrb…
Tapped out Ultra plan and a reset on some ambitious projects over the last week. 4 hours until next week. Let’s see what Astra and Argon can do on robotics.
Felt like there was some inefficiencies throughout the week with Claude. Specifically, trying to manage a complete testing system for the Robocurve project. /rc from a Raspberry Pi 5, optimizations throughout and seem to dump a lot of tokens after new testing runs. analysis later
massive personal goal seems to be on its way though: proving "Bounded Embodiments" as a software gate for all robotics: sim-only, but seems to have done its job. craigmerry.com/robocurve/#E-…
Craig Merry retweeted
Heading to Anthropic HQ to talk w security engineers and dev productivity folks on how their day-to-day work is like (+ how it's changing)
Based on the past track record, what the labs are doing, it becomes more widespread months later across startups
This was after last time:
