Off-reading
100 subscribers
475 photos
15 videos
21 files
240 links
Download Telegram
I have so many of those
6🔥4💅1
Forwarded from Cryptic thoughts
https://telegra.ph/Imagine-a-future-where-crypto-is-the-present-of-finance-How-did-it-happen-05-24
Wrote a longpost inspired by this tweet, also kept thinking of the meme attached whilst writing.

AI disclosure: fully written by me, then proofread by gpt5.5xh
https://x.com/cremieuxrecueil/status/2058647983231561938
Crazy that we witnessed the fattest humanity will ever be
🤯1
Off-reading
Photo
This media is not supported in your browser
VIEW IN TELEGRAM
NL has ~11.7 GW installed wind, about 40% offshore — roughly 4.7 GW offshore plus ~7 GW onshore across ~2,550 turbines .

Given what you’re standing in — sustained onshore Force 5–6 at the coast — the offshore farms (Borssele, Hollandse Kust, Gemini) are almost certainly at or near rated power. Turbines hit full output around 12–13 m/s, and they run at full vermogen from windkracht 6 , which is exactly today’s marine conditions. Onshore is windier-than-average too but inland speeds drop off, so figure 50–70% capacity factor there.

Back-of-envelope: ~4.5 GW offshore + ~4 GW onshore ≈ 8–9 GW of wind right now, against typical Dutch midday demand of ~15–18 GW. So yes — pumping. Probably north of half the country’s electricity is coming off that wind you’re getting sandblasted by, and day-ahead power prices are likely near zero or negative this afternoon. If you want the exact figure, energieopwek.nl or the ENTSO-E transparency platform will show it live.
🔥1
Forwarded from Cryptic thoughts
The most successful MEV bot jaredfromsubway.eth was exploited in extremely sophisticated attack for $15M. Claude estimates lifetime $60-80M PnL so not the end of the world for the team behind it, they will be vary of these new attack vectors now for sure.

Dark Forest comparison is very apt:
"It’s no secret that the Ethereum blockchain is a highly adversarial environment. If a smart contract can be exploited for profit, it eventually will be. The frequency of new hacks indicates that some very smart people spend a lot of time examining contracts for vulnerabilities.
But this unforgiving environment pales in comparison to the mempool (the set of pending, unconfirmed transactions). If the chain itself is a battleground, the mempool is something worse: a dark forest."
🫡1
"How it works: A global network of tiny seismometers
The accelerometer in an Android phone, the same sensor that flips the screen when it’s turned sideways, can also detect the ground shaking from an earthquake. If a stationary phone detects the initial, faster-moving P-wave of an earthquake, it sends a signal to our earthquake detection server, along with a coarse location of where the shaking occurred.
The system then quickly analyzes data from many phones to confirm that an earthquake is happening and estimate its location and magnitude. The goal is to warn as many people as possible before the slower, more damaging S-wave of an earthquake reaches them."

Google’s earthquake early warning system notifies users before the first seismic waves!
https://research.google/blog/android-earthquake-alerts-a-global-system-for-early-warning/
1
We’re building a digital god, labs are closing in on RSI, we’re sending data centers to space, ai gets twice as smart every 8 months while getting cheaper and now Jeff Dean quits Google to automate all research

It was a good time for me to quit, because now I’m lucky to see singularity in all of its glory. I was regretting not documenting my thoughts earlier as this is quite probably most important time humanity has ever lived through and at the very least I will be looking back at these posts.

https://x.com/jeffdean/status/2085036253263921218
🤯3
Off-reading
We’re building a digital god, labs are closing in on RSI, we’re sending data centers to space, ai gets twice as smart every 8 months while getting cheaper and now Jeff Dean quits Google to automate all research It was a good time for me to quit, because now…
I think Nov 2025 is very significant, because of claude opus 4.6 and gpt 5.2 releases, claude code was released earlier but it was the first time vibecoding really took off.
What’s the situation 9 months in?

1) Vibecoding improved dramatically: from barely usable autocompletes and copypasting from chatgpt we already arrived to many competing models, harnesses and inference providers.
All industry leaders are adopting coding agents and rapidly adapting internal infrastructure, most code releasing right now has been vibecoded.
Problems are still apparent, plenty of progress is to be made by developing additional harnesses, models are superhuman at coding, but issues like agency, niche expertise, continual learning mean that most of the code is produced by humans managing ai agents.

2) Many developments in ai safety: labs disclosed several cases of rogue behaviour, us gov is trying to regulate ai industry only 2 months after the first case with Mythos-Fable drama.
I think most population still doesn’t treat p(doom) as a real possibility but it appears that labs, governments and researchers are all treading cautiously, my personal p(doom) dropped quite a bit.

3) AI is fully mainstream, a lot of data centers water consumption psychosis, gen ai psychosis quietly moved on from "you can always tell it’s ai" to some weird hate on ai itself.
I think part of it might be unemployment which is quite bad, lots of discussions about entry level jobs

4) We no longer live in duopoly — lots of chinese os models on the frontier, os pretty much caught up to closed source models; many new promising companies in the ai market, bearish for openai, seems even worse for anthropic.

5) I saw tons of different robot prototypes, no real market so far, I’d expect we will see a lot of developments next year already, feels like enough time has passed since models became smart enough to be useful irl.

6) Lots of people even within ai bubble apparently don’t expect much gdp growth! I think people are just wrong though. "AGI would be revolutionary. It will raise GDP growth by 2-3% per year". I think vibes are very confused mostly, sentiment is all over from "ai is a bubble" (prob median position rn) to "singularity is imminent" (me tbh)

7) Saw some math researcher depressed on the timeline — ai solved a bunch of important open problems in math last few months.
Pretty much every sota model got gold on IMO, nobody is even surprised anymore.
🔥2