CyrilDeFi

vip
Active for: 1.9y
Peak Tier 0
No content yet
EU JUST FORCED GOOGLE TO OPEN ANDROID TO CLAUDE AND CHATGPT. HERE'S WHAT ACTUALLY CHANGES.
The European Commission adopted two binding orders on July 16, 2026, under the Digital Markets Act, and they took effect immediately.
Here's the actual asymmetry this fixes. Right now, downloading Claude or ChatGPT on an Android phone gets you an app. Gemini gets operating system powers the others can't touch. Invoked by holding the power button or saying "hey google." It can read your screen, act across other apps, tap into system-level functions no third-party assistant currently gets near.
The order s
  • Reward
  • Comment
  • Repost
  • Share
Coldcard didn't get hacked. it just never generated a real random number.
here's what actually happened.
back in march 2021, a firmware update meant to enable coldcard's true hardware randomness generator shipped with a bug. a supporting library only checked if the setting existed, not whether it actually worked. key generation silently fell back to a weak software substitute, seeded from the device's serial number and internal clock.
neither of those is secret. neither is random enough.
your seed phrase is supposed to come from a range of possible values so large it's effectively unguessable.
  • Reward
  • Comment
  • Repost
  • Share
All you need in 2026 are these apps to make life changing money
post-image
  • Reward
  • Comment
  • Repost
  • Share
A firmware flaw dating back to March 2021 let an attacker reproduce seed phrases by exploiting weak random number generation.
Losses are now estimated at 1,300+ BTC across 4,500+ addresses, worth close to $89M.
Long-term holders panicked and started moving funds to exchanges out of pure precaution.
My take on the saddest part.
Most people are going to walk away from this thinking “attack Trezor” or “attack Ledger” next.
Wrong lesson.
The flaw wasn’t “hardware wallets are broken.”
It was one specific firmware implementation choice, on one device, from one manufacturer.
Real takeaway: audit your
BTC-0.57%
  • Reward
  • Comment
  • Repost
  • Share
BREAKING: $BTC losses linked to the Coldcard wallet vulnerability have now climbed to $70 MILLION.
It continues...
BTC-0.57%
post-image
  • Reward
  • Comment
  • Repost
  • Share
Looking for good devs in the crypto space that want to work
Send me a dm and some examples of your work.
  • Reward
  • Comment
  • Repost
  • Share
$38 MILLION IN BITCOIN WAS JUST STOLEN FROM COLDCARD HARDWARE WALLETS. IN UNDER 25 MINUTES.
Not a phishing attack. Not a hack of the device itself. A firmware bug in the wallet's own random number generator.
Root cause: a macro that should have enabled the hardware's true random number generator was misconfigured, and a supporting library only checked whether the setting existed, not whether it was actually working. The fallback silently used MicroPython's software generator, seeded by the chip's serial number and internal clock, neither of which is secret.
Roughly 594 BTC drained from about 5
BTC-0.57%
post-image
  • Reward
  • Comment
  • Repost
  • Share
The bottom for $BTC isn't in until 53-50k range. plan accordingly
BTC-0.57%
  • Reward
  • Comment
  • Repost
  • Share
2.8 trillion parameters. 1 million token context. Native multimodal.
Kimi Delta Attention delivers up to 6.3x faster decoding at million-token context, cutting KV-cache memory by up to 75% along the way.
Attention Residuals deliver roughly 25% higher training efficiency at under 2% additional compute cost.
Built for long-horizon agentic coding and self-evolving workflows, sustaining long engineering sessions with minimal human oversight.
post-image
post-image
  • Reward
  • Comment
  • Repost
  • Share
BTC/USD - my read on the daily
O: $65,062.78 | H: $65,337.38 | L: $64,792.00 | C: $65,218.87 (+0.28%)
I'm not excited by this candle.
We're still trading under the 200 EMA ($73,696), and that line's been sloping down since April. Until that flips, I'm treating this as a downtrend, not a dip to buy blindly.
June's breakdown from the $72K–$76K zone was the real damage — that was demand turning into supply, and it dumped us into the low $60Ks fast.
What I like: price has been carving higher lows since that flush, riding a trendline off the June bottom. That's the first real sign of buyers steppin
BTC-0.57%
post-image
  • Reward
  • 6
  • 2
  • Share
GateUser-49c63007:
bullish btc
View More
this is f*cking gold
KARPATHY JOINED ANTHROPIC FIVE WEEKS AGO TO TEACH CLAUDE HOW TO IMPROVE ITSELF.
Not as a researcher who uses Claude. As the person building the team that uses Claude to accelerate its own pretraining.
His own words on joining. "The next few years at the frontier of LLMs will be especially formative."
The man who taught millions of engineers how neural networks work is now inside the lab building the next version of the model.
post-image
  • Reward
  • Comment
  • Repost
  • Share
HOW TO RUN KIMI K3, THE 2.8 TRILLION PARAMETER MODEL THAT JUST BEAT CLAUDE AT FRONTEND CODING, FOR FREE.
K3 debuted at #1 on LMArena's Frontend Code Arena, ahead of Claude Fable 5 on the benchmark that most directly measures production coding value. Six of seven domains won outright.
Here is how to actually access it right now, and what "free" really means at each tier.
The fastest free path:
Go to or the Kimi app. Sign in. No credit card required.
You get chat plus light coding assistance on the free Adagio tier.
The real limits worth knowing before you get excited:
The free tier does not un
  • Reward
  • Comment
  • Repost
  • Share
SOME USERS ARE CALLING GROK 4.5 ONE OF THE MOST NEUTRAL, TRUTH-SEEKING MODELS AVAILABLE.
There is a real study behind that claim now.
The Neutrality Project tested 18 models from 12 labs and found Grok's family sitting closest to the political center, while every other flagship model (GPT, Gemini, Claude, Llama) clustered left of it.
That said, this is a live and disputed topic.
Grok has also been the subject of extensive reporting on direct political tuning by xAI in response to Musk's own reactions to its outputs, including incidents where those tweaks produced genuinely extremist responses
post-image
  • Reward
  • Comment
  • Repost
  • Share
Fable 5 flips to pay‑per‑use in 48 hours. Burn the remaining runway to make sure you don’t actually leave anything behind.
The play: have Fable comb your repo and draft a complete skills library before the timer hits zero. Frame it as a retiring principal engineer doing a final handoff to the team, and it snaps into a totally different gear.
On the other side you get 10–16 skill files—debugging runbooks, change policies, the painful failure tales that ate days. Every instinctive judgment it would’ve made silently is now documented and reusable by Opus and Sonnet for pennies.
Single pass. About
post-image
  • Reward
  • Comment
  • Repost
  • Share
This site has tons of agents loops you can use right now
Maximize your prompting using these
post-image
  • Reward
  • Comment
  • Repost
  • Share
NOTEBOOKLM, AND OBSIDIAN TURN LEARNING INTO AN ACTUAL SECOND BRAIN SYSTEM.
NotebookLM reads and summarizes your sources, grounded only in what you actually uploaded.
Gemini expands on it, drafts from it, thinks through it with you.
Obsidian stores and connects all of it permanently, linked to everything else you have ever learned.
Three tools. One loop. Knowledge that compounds instead of disappearing the moment you close the tab.
post-image
  • Reward
  • Comment
  • Repost
  • Share
ONE PROMPT THAT FORCES YOUR CLAUDE AGENTS TO ACTUALLY THINK INSTEAD OF JUST ANSWER.
Act as a world-class analyst and operator.
Before answering, identify hidden assumptions, missing information, and potential failure points in my request. Then provide:
(1) the best answer
(2) an improved version of my question
(3) alternative approaches
(4) what experts in the field would likely disagree about
Be concise but thorough.
Most prompts ask for an answer. This one asks for the thinking behind the answer first.
The improved question alone is worth running this prompt for. It surfaces exactly what you
  • Reward
  • Comment
  • Repost
  • Share
  • Pinned