Mats Erdkamp (@mats_zip) header
Mats Erdkamp (@mats_zip) profile picture

Uses lightning infused rocks to center divs all day. DMs are open • Design Engineer at @thirty_five_com • Industrial Design Graduate

The Netherlands
Joined November 2011
Photo by Mats Erdkamp (@mats_zip)Photo by Mats Erdkamp (@mats_zip)Photo by Mats Erdkamp (@mats_zip)Photo by Mats Erdkamp (@mats_zip)Photo by Mats Erdkamp (@mats_zip)Photo by Mats Erdkamp (@mats_zip)Photo by Mats Erdkamp (@mats_zip)Photo by Mats Erdkamp (@mats_zip)Photo by Mats Erdkamp (@mats_zip)Photo by Mats Erdkamp (@mats_zip)
Recreated an image by @rauchg and made it dynamic with @threejs Link to try it out in the replies!
16
16
3
445
25,971
why arent all heat sources everywhere just a GPU
31
Started measuring project in FTAs (full-time agents)
1
55
Mats Erdkamp retweeted
Not sure when but I’d love that
1
1
118
Bro was ahead of the game
1
1
1
2
114
Mats Erdkamp retweeted
Replying to @skooookum
1
76
15,318
“In late spring 2020, a source I knew to be reliable called me with a story,” writes Matthew Cole. “The Dutch government had hacked into Jared Kushner’s phone and monitored him as he accessed a Dubai-based bank account. What was odd, the source said, was that the money in the account—hundreds of millions of dollars—had come from a bank in Azerbaijan. Most importantly, Kushner appeared to have transferred significant amounts of money from the Dubai account to other individuals, including a warlord, a mercenary and a one-time hit man. The story sounded preposterous, I told my source. And yet. Six months later, I found myself speaking with someone else about the UAE government and their use of mercenaries in conflicts around the Middle East and Africa when I asked, on a whim, whether they knew anything about Kushner and bank transfers from a UAE account. My acquaintance grew quiet. Then they replied, ‘Yes.’” Kushner’s lawyer denied the bank account existed. Read an exclusive excerpt from Cole’s book, ‘Princes and Thieves: Jared Kushner and Mohamed bin Zayed’s Secret Plot to Remake the Middle East’: vanityfair.visitlink.me/Q5_W…
1
65
Mats Erdkamp retweeted
--- BEGIN PRIVATE KEY ---
39
765
17
13,274
158,843
Mats Erdkamp retweeted
interpretability might be the most important scientific problem in the world right now, as I don't think we can solve alignment if we don't know what we're building. I think I have good news about it. I co-authored early work in mechanistic interpretability at OpenAI and came away pessimistic, because neural networks appeared to fold into arbitrary tangled knots. years later, I started openai's first large-scale automated interpretability project, finding about 2% of neurons legible with the primitive tools we had with modern agentic tools and a much closer look, the picture looks totally different. the vast majority in models up to trillions of parameters are legible, and their circuits beautiful. many famous examples of tangled neurons turned out to be misunderstood. The roadblocks I expected mostly aren't there. I now think ambitious interpretability will be solved by the time we need it most, and possibly quite soon, as automated scientists keep improving. it may be a bitter lesson for interpretability: the simple thing, at absurd scale, works already-clean.goodfire.com/
885
2,637
1,446
19,860
4,572,776
My body is a vessel converting Cloudflare birthday week posts to Claude prompts
2
3
2
22
1,861
But will it be able to write good javascript abstractions
Jacob is correct here—we really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade. I believe Anthropic is trying its best, but we do not yet have a plan to solve alignment for superintelligence and are not clearly on track to.
139
yo @mattzcarey @threepointone I had a crack at adding speculative programmatic tool calling, by the goat @a1zhang, in the codemode runtime. I think it's a really cool idea, and it seems to work nicely with CF agents and the RLM example. Alex’s post: alexzhang13.github.io/blog/2…
144
QUICK SOMEONE SEND HIM SOME BANTER
I asked Codex to pull up some stats and I receive on average one DM or email every 6 or so minutes to ask for a reset. I occasionally do oblige if it comes with really solid feedback or banter.
1
177
Mats Erdkamp retweeted
Build something. Put it out into the world, then watch what others create with it. There’s nothing more satisfying than that.
Free idea for someone to iterate on: turn @evilrabbit_'s excellent lifeline into a learning tool. I added MCP and asked it to give me a timeline on the history of compute. Needs refinement but potentially quite useful! Could also work nicely with @mattpocockuk's /teach-me skill
2
3
73
9,414
Free idea for someone to iterate on: turn @evilrabbit_'s excellent lifeline into a learning tool. I added MCP and asked it to give me a timeline on the history of compute. Needs refinement but potentially quite useful! Could also work nicely with @mattpocockuk's /teach-me skill
2
1
1
21
10,293
Mats Erdkamp retweeted
. @KentonVarda was just ahead of his time again. There is still work to be done to make the technology easily by the rest of us. Capabilities security models are just not the mental model most people have for security.
Replying to @dok2001
The return of sandstorm io 🥹
3
1
63
10,854
Mats Erdkamp retweeted
hello there the jacobian conjecture is false thanx to my close friend akhil for asking about it and my other close friend fable for working during the world cup final ((1+xy)^3 z + y^2 (1+xy) (4+3xy), y + 3 x (1+xy)^2 z + 3 x y^2 (4+3xy), 2 x - 3 x^2 y - x^3 z): \C^3\to \C^3, has jacobian determinant -2, and sends (0, 0, -1/4), (1, -3/2, 13/2), and (-1, 3/2, 13/2) to (-1/4, 0, 0)
1,699
5,380
3,151
43,795
40,102,165
Odysseus learning of his evil ways not through divine punishment, but by 500 lil yellow dudes showing up
If the minions were Odysseus’s crew shit would’ve gone down differently
204
Mats Erdkamp retweeted
I wish the jamaican man in yeezus had more parts
47
2,466
406
6,618
Tfw your /goal mode run suddenly completes
1
144
Long term prediction OpenAI buys Microsoft
1
2
157