Luca A.K.A Pulp Crypto (@PulpCrypto) header
Luca A.K.A Pulp Crypto (@PulpCrypto) profile picture

Crypto since 2020 | AI, infra & markets | I check things before I believe them | Oldschool @DeFi_Italia

Milan, Italy
Joined April 2025
Photo by Luca A.K.A Pulp Crypto (@PulpCrypto)Photo by Luca A.K.A Pulp Crypto (@PulpCrypto)Photo by Luca A.K.A Pulp Crypto (@PulpCrypto)Photo by Luca A.K.A Pulp Crypto (@PulpCrypto)Photo by Luca A.K.A Pulp Crypto (@PulpCrypto)Photo by Luca A.K.A Pulp Crypto (@PulpCrypto)Photo by Luca A.K.A Pulp Crypto (@PulpCrypto)Photo by Luca A.K.A Pulp Crypto (@PulpCrypto)Photo by Luca A.K.A Pulp Crypto (@PulpCrypto)Photo by Luca A.K.A Pulp Crypto (@PulpCrypto)
Starting a series: the AI Release Radar Every two weeks, a ±45 day sweep on what already shipped and you missed, what has a date and what's real work with no date attached. Not a rumor list Let's dive into Issue 01 👇
2
1
6
259
We started out hiding, persecuted and mocked but now we have likely launched the revolution that will lead to ALL markets being open 24/7 You can't stop innovation
JUST IN: 🇺🇸 SEC prepares "innovation exemption" to allow 24/7 blockchain trading of tokenized stocks.
1
1
104
Grok 4.6 benchmarks ore out and they look amazing
2
60
I just noticed while scrolling the CT that NFT's are back in vogue and, surprise, there are familiar faces behind them (who would have guessed?). SpoiIer: If you still let yourself get ripped off like this in 2026 no 1T parameter frontier model will save you from going rekt.
1
2
53
Mine cancels yours so yours will cancel mine back? The agent that found the hole in that Australian gym API was not malfunctioning: it did the job well and one agent doing the job well is great news. Thousands of them doing it well against each other is beautiful engineering and a real problem: prices bid up by mutual overbooking, APIs falling over from the traffic and what else?
A man in Australia asked his agent (Claude running on OpenClaw) to book him a spot in a popular gym class. The agent found a software vulnerability that let it book the class weeks further ahead than should have been possible. When the user then asked if it could move him up the waitlist, the agent discovered the API had no authorisation checks on cancelling other people’s reservations, so it cancelled the person in the first spot and moved him up the list. Some people will call this misalignment, but his agent was perfectly aligned to him - it was only trying to help its user get what he wanted. The most important thing about this story, in my opinion, is that it gives you a window into what is about to start happening on a massive scale once millions of people have an agent trying to get their beloved users the best seats, bookings, appointments or reservations through absolutely any means necessary.
1
477
Sixteen people pointed AI models at Bitcoin open source and came back with 4,962 findings, 85 of them critical. Far fewer will stand once professionals recheck the list and they have every reason to hurry. Nobody pays the maintainers and their work is the valuable part: discovery starts here but whether volunteers make the number honest or just unfiltered we can't call yet.
1
1
47
Everyone's quoting Hyperliquid revenue down 43% Ok cool, but that's measured against Q3 2025. Quarter over quarter it went up, $185.5M to $202M The number that moved: volume routed through HIP-3 markets went from 20.7% to 32.2% in one quarter. A third of the flow now shares fees. Holding HYPE gets less rewarding while the OI chart stays up there and says everything's fine.
1
2
236
OpenAI’s frontier model Astra did two things in a six days timeframe One cost $2k, the other forced them to hit the brakes Same capability Thread 👇
2
50
Curious how many people already saw the math announcement versus the safety slowdown 🤓
11
3. August 7th: OpenAI slows Astra development. Preliminary evals couldn’t rule out “Critical” cyber capability. Autonomous zero day exploits on hardened real world systems. First time one of their models reached this level
1
21
4. Same skill set: long horizon search where a machine can instantly check if the answer is correct. Works for theorems, works for exploits. Capability shows up first, governance tries to catch up. AI testing and safety keeps looking like one of the more concrete near term sectors.
8
Many are selling "miracle" prompts on CT but the real power of prompting lies in adapting and optimizing for your specific needs, not generic requests Anyone selling you ready made prompts is stripping you of your advantage
1
42
I find hard to see how they could realistically succeed (okay, you win but then what?) but I see yet another pattern emerging in the West: private capital stepping in to do what governments neglect whether it’s the race for chips, technological development or even upholding justice. Capitalism at its best.
🇰🇵 LATEST: Bybit has sued North Korea and the Lazarus Group over the $1.5 billion hack and has secured a preliminary court order freezing stolen assets.
60
Ten years ago AlphaGo played Move 37: a move that no human would have made (1/10000 odds) Behind it years of work by Hassabis and the DeepMind team: neuroscience, game theory, reinforcement and the goal of building systems that could discover strategies no human had mapped yet
3
58
Sources: • WSJ “Move 37 Is the Moment AI Changes Everything” (Ben Cohen, Aug 8 2026)
25
It wasn’t just raw computation That moment marked the real shift : AI stopped merely replicating and started exploring Today, as the WSJ notes, those “Move 37” moments are no longer rare, they’re becoming the normality
1
18
The point of no return wasn’t when machines started winning, It was when they started surprising us everyday Would you like an in depth thread about the "Move 37" incident? Comment below 👇
16
JUST IN : ByteDance is cooking a ~10T parameter model. SoftBank just booked a clean $2.2B mark-up on its stake and the only real crypto bridge still stands: BytePlus × Sui. Here’s the full picture ↓
2
1
83