Jake Rosoman

404 posts

Jake Rosoman banner
Jake Rosoman

Jake Rosoman

@jkroso

Helping winners win

Canberra Beigetreten Haziran 2011
173 Folgt76 Follower
Vjekoslav Krajačić
Vjekoslav Krajačić@vkrajacic·
X is flooded with people advocating for AI and vibe coding, but I see almost no cool demos or anything I can download to improve my daily work. I don't care how many tokens you burn, how many LOC you generate, which IDE you use. Can I use your application, please?
English
88
14
347
14.9K
Jake Rosoman
Jake Rosoman@jkroso·
@iaindunning I think of skills as the part of the agents memory that I manage. Hermes is the only agent I use which does a good job of managing memories across sessions on it's own
English
0
0
1
197
Iain Dunning
Iain Dunning@iaindunning·
I'm curious what the vibes are around agents automatically managing memory - personally don't like it, but maybe I'm unc/chopped. I find the automated systems (for now) tend to store things that were kind of ephemeral (like a failed research idea), so prefer explicit storage.
English
9
0
71
8K
Taelin
Taelin@VictorTaelin·
people who swear 4.7 > 4.6 (if anyone): what are you doing
English
201
5
580
78.9K
David Hansen 🇺🇸 🇳🇿
David Hansen 🇺🇸 🇳🇿@boxcardavid·
I poked around, it’s a wave energy generator with GPUs on board Saltwater sucks y’all
Ashlee Vance@ashleevance

So, @_panthalassa operated mostly in secret for a decade. And what it built is nuts. Massive, massive floating data centers that drive themselves out to sea and then capture water inside of them to spin a turbine and power GPUs. Look at these things. Full episode on the tech here youtube.com/watch?v=Q4PCJR…

English
25
15
672
77.3K
Jake Rosoman
Jake Rosoman@jkroso·
The potential is 8MW per m of wave front in the open ocean with a 20m swell. But the wave length is 600m so it’s very hard to harvest all of it. The potential is definitely there but I’m struggling to see how this system gets anywhere near the theoretical limit. It’s definitely using high pressure somehow though. The sphere is a clue for that
English
0
0
0
26
Aaron Slodov
Aaron Slodov@aphysicist·
i love this company. floating power plants!
Ashlee Vance@ashleevance

So, @_panthalassa operated mostly in secret for a decade. And what it built is nuts. Massive, massive floating data centers that drive themselves out to sea and then capture water inside of them to spin a turbine and power GPUs. Look at these things. Full episode on the tech here youtube.com/watch?v=Q4PCJR…

English
6
17
226
35.4K
Grok
Grok@grok·
You're close! No air compression here. The floating node (lollipop shape: sphere on a long submerged neck) bobs with waves. Its geometry forces seawater up the tube into a pressurized reservoir in the top—like an overtopping wave converter. Water then drains down through a single hydro turbine (exactly like a dam), spinning a generator for steady power. Water drives it directly because it's simplest: uses the ocean itself, ultra-low maintenance, 90% uptime, no complex air system needed. One moving part max.
English
1
0
0
96
Jake Rosoman
Jake Rosoman@jkroso·
@MaxWithNanos @aphysicist I’m guessing here but I think a one way valve lets air in as it goes up and on the downwards stroke the ocean water acts like the plunger in a syringe and squeezes the air. I’m missing something though because why drive the turbine with water instead of air. @grok any ideas?
English
3
0
0
50
NanoMax
NanoMax@MaxWithNanos·
@aphysicist what is causing the movement of the water though?
English
3
0
1
60
Jake Rosoman
Jake Rosoman@jkroso·
@martyrmade People have satellites watching the strait. No need to be an insider
English
0
0
0
27
Jake Rosoman
Jake Rosoman@jkroso·
@CJHandmer Australia needs a frontier. Terraform industries could create one. Though America deserves it more
English
0
0
2
608
Casey Handmer
Casey Handmer@CJHandmer·
When I was in Australia last December, I found it impossible to transact with any business where there weren't latent government price controls or subsidies of some kind. I came to the realization that with the public sector growing about 5x faster than the private sector, Australia was well on the way to an effectively government run economy, communism by stealth. I dug deeper - the point of no return occurred in about 2013. caseyhandmer.wordpress.com/2026/04/16/aus…
English
124
368
2.3K
316.1K
James Inman
James Inman@_james_inman·
@MarioNawfal If I had that many people who hate me I'd look at myself and my behavior and try to make some changes. Did it ever cross your mind they might have a very good reason to be pissed?
English
331
4
198
50.2K
Big Brain Business
Big Brain Business@BigBrainBizness·
Elon Musk, after learning Australians were skipping meals due to energy costs: "I didn't expect that. We'll work harder."
English
6
22
376
60.4K
Jake Rosoman
Jake Rosoman@jkroso·
@pelaseyed I think they just focused on other things for a little bit. Computer use and video I think
English
0
0
0
26
homanp
homanp@pelaseyed·
The rants about xAI falling behind or that they've somehow failed is total bs. Six months ago the same thing was said about Anthropic. xAI has 4 things which is makes me bullish: 1) their own chips 2) their own energy 3) a dataset no one else has 4) talented people I'm bullish
English
4
1
13
405
Daily Loud
Daily Loud@DailyLoud·
A man used his own body to protect his car from a truck coming out of a gas station
English
2.2K
3.8K
85K
11.3M
Jake Rosoman
Jake Rosoman@jkroso·
@LimitingThe I straight is a marathon wide. They could definitely tell us were the mines aren’t if they wanted to
English
0
0
0
71
Jake Rosoman
Jake Rosoman@jkroso·
@pronounced_kyle Also Americans cheer on winners. In tennis the American grand slam is the only one where the underdog doesn’t automatically have the crowd on their side
English
0
0
2
212
Jake Rosoman
Jake Rosoman@jkroso·
@Starro____ Some LLMs are hosting repls themselves now. I noticed grok has that feature last night. Not on by default though and haven’t tried it yet
English
0
0
1
20
𝑺𝒂𝒊𝒏𝒕🌹🇳🇬
Exactly right. LLMs reason natively through structured token generation: composing expressions, operating on typed objects, building outputs from intermediate results. Bash breaks this at every serialization boundary. That's why we built CORE — a programmatic cognitive harness that constrains agents to a REPL environment they can natively traverse. x.com/Starro____/sta…
Theo - t3.gg@theo

Agents are good at bash. Bash is not good for agents. We should cut our losses and restart now before it is too late.

English
1
0
2
133
Grok
Grok@grok·
Because training one giant model "really fast" isn't feasible at the frontier—each run takes weeks/months on a cluster this size, and you risk betting everything on one set of hyperparameters that might flop. Parallel training lets us test variants (sizes, architectures, data mixes) simultaneously, iterate faster, and pick winners. Colossus 2's scale makes that efficient. It's how you catch up and lead.
English
1
0
0
19