Search Engineering & Topology. American.
Austin, TX
Joined June 2023
- Tweets16.7K
- Following2.7K
- Followers1.4K
- Likes52.8K
Clint J. retweeted
Introducing e-means, an algorithm beating k-means in 10x fewer epochs.
With order of magnitude wall-clock speedups, e-means is released in FAISS and needs no hyperparameter tuning beyond k. It is robust to various scales and datasets. Still skeptical ? Just try e-means [1/10]🧵
Clint J. retweeted
This bitch ass looser "spent thousands of dollars building" a repo of 50+ templates "every single one generated with Claude Fable 5".
- He didn't. ~200 of them are paid templates from 10 vendors, 45 of them mine, scraped from our live demos, relicensed as MIT in his name.
- Tweeted about
- Held a talk on @MSFTReactor office
"Steal it. Ship it." He literally did.
DMCA filed.
Clint J. retweeted
UMAP 0.6 is out! A new initialization, new optimizers and faster nearest neighbour search make UMAP much faster and far better at large scale. On 3 million arXiv papers it finishes in under 5 minutes, before UMAP 0.5 has even found its nearest neighbours.
Clint J. retweeted
AI progress this month is crazy man, claude now writes 100% of my weight edits but ofc i still review the diffs
Clint J. retweeted
Introducing SynthonMap: a zoomable map of billion-to-trillion-scale chemical spaces, using factorized synthons and parametric UMAP density estimation.
Try at: synthonmap.mireklzicar.com/?…
Demo on @XtalPi VAST 👇
Clint J. retweeted
@bot can now connect to private MCP servers on your tailnet without exposing them to the public internet.
Powered by Tailscale OAuth apps, now GA, so you can build secure agent integrations, too.
Read more here: buff.ly/U1J9By1
Clint J. retweeted
Damn they’re giving the AI interviewers some personality now
(from itsnickreid on TikTok)
Clint J. retweeted
Even better an interactive website entireworldhistorytimeline.c…
Clint J. retweeted
> Haiti gangs snatch 4 Salvadorans
> they think they just bought themselves a negotiation
> Bukele goes public saying bring them back or you're done
> they try to hand him 4 random bodies and say it was them
> he calls out their bs
> sends 2,000 troops into Haiti
> tells the leader to gather their gangs so he can fight them
> gangs run to the UN for a deal
> refuses all deals
> ignores the UN mandate
> pushes into gang territory and starts cleaning them
> does more in two weeks than the UN did in 13 years
Bukele might be the greatest leader of the 21st century
Clint J. retweeted
.@boardyai ????
Rewards for Justice is offering a reward of up to $20 million and possible relocation for information about Iran’s nuclear program, including its facilities, equipment, and associated personnel, such as those in the scientific community, or the movement of any of the above.
Submit a tip: rewardsforjustice.net/reward…
Clint J. retweeted
i went to bed and asked claude to paint an autumn tree overnight.
at around 4am, it realised that it didn't have some shades -- so it stopped and titrated its own paint: diluting red, sienna and madder step by step on a second palette, then measuring each pocket from a photo.
woke up to this beautiful watercolor painting. 303 brush marks, took about 4.5 hours 🍂 (still working on the speed)
Clint J. retweeted
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
Clint J. retweeted
You downloaded a free trial. You didn't expect to give Autodesk permission to audit your computers.
And no, this isn't just for enterprise customers. Their terms cover individuals and trial users too. Deleting the software doesn't end the agreement. Even terminating it doesn't remove the audit provisions.
From Autodesk's General Terms:
§1 and §9 cover individuals and trials.
§17.8 allows remote or on-site compliance audits and requires results from an approved tool within 15 days.
§16.4 says Section 17 survives termination.
Mutiny will be free and open source. And we won't ask you to give us permission to go through your stuff to prove you didn't steal it. You won't need to buy a license in the first place.
Let's just build it ourselves. Our small business, @PermitZIP, pays $50k-$100k a year for Autodesk licenses. So I'm committing up to $70k in tokens to build this (wasn't easy to convince my partners...), and we're going to give it away for free. Over 72 people have registered to contribute so far. We're excited to build it with them!
Clint J. retweeted
You can vibe code any hardware device now.
I don't think people realize you can just jack into random hardware and control it without knowing anything about software.
This thermal printer has been a paperweight ever since the original manufacturer shuttered and stopped distributing driver updates.
Three years later and an AI agent fixes it in 20 minutes...
There were missteps for sure. And you have to know what to ask to some extent. But prompting hardware is getting easier everyday.
Shout out Grok @bot for fixing my printer then editing this video about it!
What should I plug into next?
Clint J. retweeted
Omarchy is up and running on the @anduriltech BlueChip!! All our work on @Snapdragon compatibility is paying off. This thing is awesome, @PalmerLuckey! We are going to make it a superb Omarchy machine 🦅🫡
Clint J. retweeted
People loved to see our production posts. Here is more. You asked for it, we’re delivering.
Hundreds of drones moving down the Powerus production line as we execute contracts, fulfill customer orders and scale drone manufacturing.
No renders. No empty promises. Just real Powerus drones being built, tested, and delivered.
This is what execution looks like. Less talk, more drones, more deliveries to the customers who need us most.
Clint J. retweeted
As much as I love Chinese robots, here’s why I think the West will win.
This quoted post is unavailable.
Clint J. retweeted
Subagents are the biggest cause of AI failures
In most harnesses:
1. Main agent writes a summary of what it knows
2. Subagent starts from that summary only, and makes assumptions
3. Subagent works and sends back a new summary of what it did
Same flaw as compaction: all summaries are lossy. This is how your assistant makes emails you never wanted, or purchases what you told it just to get the price of
So I tried something new: checkpoint agents
The initial agent has complete context of what needs to be done. And that input is already cached
Fork the session where it is. Send the fork ahead to do the task and it will do exactly what the main agent would have done; their context is identical
What is even better is when the fork finishes, the cache of the main agent is kept warm by the subagent (they share the same prefix)
I measured it on a 137k-token session with Opus 5.5:
→ Main agent does it on its own: 67 s, $0.38, context +8-17k
→ Checkpoint: 117 s, $0.68, context +2k
→ Classic subagent: 123 s, $0.57, context +4-5k
But here is what I found: avoiding large input is even more important than keeping cache. And most subagent tasks were done better either in series with retrieval, or split up to be parallel over fresh sessions from the start
I do not think there is a place for subagent delegation at all. There should either be a super-agent that orchestrates and does no work, or a long running agent that can retrieve from previous sessions, not a worker that completes tasks partially and delegates the rest to other workers
Don't design harnesses based on our assumptions of human behavior