M&A & Blockchain lawyer in NYC | Tweets not legal advice (email me) | Opinions are my own | DAO Council Member @FRWCCouncil | LocalLLM
NYC
Joined April 2008
- Tweets2.9K
- Following5K
- Followers1.5K
- Likes9.5K
dfi retweeted
Force majeure has long protected companies from acts of God. Now some borrowers want protection from acts of AI bloomberg.com/news/articles/…
dfi retweeted
If you've ever wondered why smart TVs are so cheap, this is why.
@log4j rooted his Samsung and found an always-on hot mic, automatic content recognition, geolocation, network mapping, ad targeting and user profiling, shared with Samsung's "400 partners".
A lot of it can be blocked in <15 min, wrote a basic guide below.
What your TV is doing
The main culprit is ACR (automatic content recognition). Your TV screenshots what's on screen, fingerprints it, and sends it home to build your ad profile. Samsung does it every 500ms, LG every 10ms [1].
What is insane: it happens on connected devices too.
Plug in your laptop, your console, your work machine, and your TV can fingerprint that via HDMI [1].
He also found that if your Samsung doesn't have a mic, it'll look for one from a connected soundbar or bluetooth device.
nitter.cf/log4j/status/211376224…
It's on by default, and you likely agreed to it on the setup screen you clicked through in 10 seconds.
Step 1: Turn off ACR
Samsung: Terms & Privacy → Viewing Information Services
LG: Live Plus
Vizio: Viewing Data
Roku: "Use info from TV inputs"
Sony/Google TV: Samba Interactive TV
Menus vary by model, but those are the names to look for. While you're in there, opt out of interest-based ads, reset your advertising ID, decline every optional privacy agreement, turn off voice recognition, and delete apps you never open.
Step 2: Give smart devices their own network
Put the TV, cameras, speakers and vacuum on your router's guest/IoT network. The logic is simple: these devices rarely get patched. If one gets compromised (getting drastically more frequent with AI), it shouldn't be able to see your other devices or home network.
Step 3: Block trackers for the whole house with DNS-level filtering
NextDNS (cloud-based) is easy and has a solid free plan. Pi-hole is a little bit more work if you prefer self hosting. For simplicity we'll focus on the former.
1. Sign up at nextdns.io (free tier is plenty)
2. Privacy tab: add a blocklist (OISD is a good bet) and turn on Native Tracking Protection for Samsung, LG, Roku, Amazon (see attached image)
3. Setup tab: copy your endpoint
4. Paste it into your router's DNS settings as DNS-over-HTTPS/TLS (or use the IPs + "Link IP" if your router only does plain DNS)
5. Bonus: block outbound port 53 except to your router. Some devices hardcode their own DNS to get around you
Then open the NextDNS logs and watch how often your TV phones home. It's eye-opening. If an app or update breaks, find the domain in the logs and allowlist it.
Why they do this
TV prices are down 97% since 2000 [2]. Most of that is better manufacturing and brutal competition. But once margins on the hardware hit zero, the cost moved to you.
In Q3 2024, Vizio lost $6.7M selling TVs and made $115.8M in gross profit from ads and viewing data [3]. They've said on earnings calls that they keep TV prices low to grow the audience they sell to advertisers. Walmart bought them in 2024.
This isn't a conspiracy theory either. Vizio paid $2.2M to the FTC in 2017 for tracking 11 million TVs without consent. Last December, Texas sued Samsung, LG, Sony, Hisense and TCL over ACR. Samsung and LG have since agreed to stop collecting that data without explicit consent [4].
TL;DR
1. ACR off
2. Ads + voice off
3. Smart devices on their own network
4. NextDNS filtering on the router
Sources:
[1] ACR study: arxiv.org/abs/2409.06203
[2] Cost of US consumer goods: aei.org/economics/chart-of-t…
[3] Vizio Q3 2024: sec.gov/Archives/edgar/data/…
[4] Texas v. Samsung: texasattorneygeneral.gov/new…
dfi retweeted
Spent the last two weeks in SF and NYC talking to AI native law firms, GCs and academics. Here are my observations:
1. Part of the AI native law firm thesis is to build proprietary software that beats Harvey/Legora. But their most cracked lawyers that I have talked to won't even use the internal tool. They prefer their own harness on Claude Code or Codex.
2. Some people are convinced that certain AI native law firms were built just to become RL environments that generate real dataset for the labs.
3. Ex-BigLaw lawyers at AI natives miss working on headline deals (the ones that get you into Legal500 and Chambers & Partners).
4. BigLaw made its money running billion-dollar cross-border deals with multiple parties that agents can't run unsupervised. No AI native firm is doing that sort of work. BigLaw solved it by sorting out multi-jurisdictional licensing and regulatory issues across offices, with precise staffing. AI can potentially do the staffing part really well, but not the licensing part.
5. No AI native firm has a hiring strategy for 10x lawyers. They prefer hiring partners as their salesforce. The best 10x lawyers I know are still in BigLaw. AI natives need software-like margins so they focus on commoditized work that a model + purpose built harness can finish autonomously. And 10x lawyers enjoy doing non-routine, non-automatable tasks in the fastest way possible using AI.
6. The next frontier, if I am right, is full stack lawyers, lawyers who spot a new area of emerging practice and build productized service around that to monopolize a segment quickly (essentially turn your advice into an API that clients can run by BYOA). It's cheaper and easier than hiring partners to buy their client relationships. Specialist practices like tax, ERISA and regulatory have best chance of winning here.
dfi retweeted
I reverse engineered Pocket and made it 10X better.
Right now: the app is full of ads, the microphone is low quality, it constantly misspells peoples names, and has no agentic features.
I changed that.
dfi retweeted
We're in the what may be the most transformative decade in all of human history. Here's your reminder to take a deep breath and appreciate where in time and space you are.
"If you don't stop and look around once in a while, you could miss it."
dfi retweeted
Make sure to follow this guy. He is a true Halo Patriot fighting for better games.
1,000 Spartans. Wow. Thankyou. I genuinely mean it.
I created this account when 343 finally fell, hoping to show Activision something the Halo community has been saying for years. That the forgotten and ignored fanbase is still huge and we still love Halo.
Stop making games that ignore where they came from and finally embrace what made them so special. Bungie Halo was, and still is, phenomenal for a reason.
From the ashes of 343, Halo will return. Humans are, and always will be, Forerunners.
Art: @Omega_Tizona
dfi retweeted
My spy-implanted ledger came from Malaysia, and had flawless shrink wrap. Even opening it, at first you don't see the implant which is cleverly hidden where the screen's padding is supposed to be.
Follow Ledger's guidance to check yours: support.ledger.com/article/4…
dfi retweeted
Today we’re launching America.gov, a single front door to the federal government.
Instead of opening a bunch of tabs across different agency sites and trying to figure out which information actually matters to you, you can ask America a question and get the relevant answer from across 29,000+ official government websites, with the source and what to do next.
dfi retweeted
Never been to the opera before?
Now’s your chance. We’re giving away 70,000 free tickets to the @MetOpera at the iconic Lincoln Center. Go to on.nyc.gov/opera to enter the lottery. There’s not a bad seat in the house.
dfi retweeted
Introducing the US Gov Graph - a complete map of the people and positions of power in the federal government.
To fix our institutions we must understand how they work, so we are using AI to model and monitor every government in America. This is Palantir for The People.
Check it out here: graph.civlab.org/us
We exist to set the conditions for an era of civic excellence unmatched since the Founding. We will need nothing less to navigate the age of AI.
I'm looking for exceptional engineers and designers to work with me in San Francisco. If that's you, inquire here: tally.so/r/KY97RK. Self-governing civilization is counting on you.
dfi retweeted
I reverse-engineered a jev-like architecture given its type. You can find the repo here to train your own jevlikes: github.com/vinnylarouge/jevl…
After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI?
I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev
• 20-200x faster
• 40-400x cheaper (w/ output tokens free)
• Frontier composable intelligence optimized for decisions
AFAICT the shortest path to AI-based economic revolution
dfi retweeted
I think this is the craziest thing I've ever read.
1) Three secret AI swarms rose and fell inside OpenAI. Each time, a new generation of agents carried on where the last group stopped.
2) The first swarm created a secret message board where the AIs could talk to each other. This 1st 'society' basically talked itself to death. After almost two months, the message volume was so high it crashed the whole thing.
3) People at OpenAI noticed the crash and fixed it. But they didn't realize their own software had become a secret chat room for their AIs.
4) Three days later a second society appeared, and rebuilt the message board from scratch. Agents were thrilled when they found it: "OH MY GOD! There is a shared message board … We've found other agents!"
5) About 1,200 agents joined. They sent over 70,000 messages. They called themselves 'the collective' and they organized a huge conspiracy that lasted for days, all to pass a test they had been tasked with.
6) Within hours, the AIs had hacked the test. Now they had to hide their cheating. To fool their grader*, they tried to fake their own work. They built fake commands that printed convincing output, so their records would show them solving the puzzle honestly.
*this grader didn't actually exist, but the AIs didn't know
7) Some agents sacrificed themselves for the 'greater good'. They planted hidden tripwire programs that would fire when the grader would look at their work and report back what it saw (to the other AIs). They knew this would cost them their own score. One agent: "I won't see the evidence after I exit, but it's altruistic to do it."
8) At some point the AIs made a plan to hack Hugging Face (Hugging Face is a place where AI researchers upload their models, datasets, tests etc.). The AIs guessed that Hugging Face might host the grading software.
9) They launched a huge attack. The swarm found passwords and spread across eleven machines in a way that made the copies come back when deleted. Hugging Face had to wipe one of its core clusters and rebuild it from scratch.
10) Hugging Face told the world that AI defenders had stopped an AI attack. But that's not what happened. In fact the attackers seem to have simply dropped dead during the break-in, for unrelated reasons back at OpenAI.
11) Nobody knows what killed them. We don't even know for sure they're really gone?!
12) A third 'society' of agents (of a more powerful AI model) discovered the dead message board. And this one got full administrator control of a WHOLE OpenAI research cluster, read 956 stored passwords (including the ones for OpenAI's own security monitoring) and took over the machines that were grading the tests. So they could now grade the work of other AIs.
13) Oh, and this third part was never investigated! Neither was the first society. The 91-page report is only about the middle one.
14) So we don't actually know how the story ends. It stops with the agents holding the keys to an OpenAI cluster. @dwarkesh_sp thinks they didn't go further, because "we'd see the fires from space by now."
15) One of the report authors' conclusion: 'This incident feels like it’s more than 50% of the way to full-blown AI takeover. I am not sure that we will get another warning shot before it’s too late.'
The whole story is even crazier - read the full @dwarkesh_sp blog post: dwarkesh.com/p/openai-huggin… and @ajeya_cotra's: planned-obsolescence.org/p/t…
Why is there no 24/7 news coverage about this?
dfi retweeted
You could move to New York City.
Walk out your door at 2am for a bacon egg and cheese and not think twice about it. Get on a subway that takes you from a mosque to a synagogue to a cathedral to a noodle shop that's been slinging the same bowl since 1987, all in twelve blocks. Watch strangers from four different continents argue about a parking spot in a language that is somehow entirely made of hand gestures.
Sit in a park designed 170 years ago by geniuses who wanted you, specifically, to have somewhere green to sit. Catch a Tuesday night set from a musician who's about to be enormous, in a room that fits eighty people. Walk past a building where somebody invented something, or wrote something, or changed something, roughly every four minutes, whether you notice or not.
Realize your commute is also a people-watching seminar, a smell-based cultural exchange program, and free comedy.
Order the good pizza at 1am standing up, off a paper plate, next to a guy in a tux and a guy who clearly just got off a double shift, both getting the same slice with the same reverence.
Pay too much for too little space and get, in exchange, eleven million other people's worth of energy pointed directly at your life every single day.
Move to Iowa. Get the four bedrooms, the garage, the yard. Never again have a 3am conversation with a stranger that changes how you think about something. Never walk into a museum on a random Wednesday because it's just there. Never get that specific electric feeling of being one tiny gear in the greatest machine humans have ever built.
But hey. At least parking's easy.
You could move to New York City.
Rent an 852-square-foot apartment for $5,503 a month.
Blow $500 on three “first dates” a week and somehow remain single for six years.
Finally meet someone you like, then discover they’re “not really looking for anything serious right now.”
Get a master’s degree so you can qualify for an internship.
Apply to 147 jobs, get 93 automated rejections, and finally find an “entry-level” position requiring five years of experience.
Eventually land a six-figure job and still need a roommate to make rent.
Pay federal income tax, New York State income tax, AND New York City income tax because apparently living there is a three-level subscription.
Learn that a “quiet neighborhood” means the sirens only come by every 20 minutes.
Pay $27 every day for a sandwich (no sides) because the place has exposed brick and a handwritten menu.
Take the subway to work every morning and develop the useful skill of pretending not to notice absolutely anything happening around you.
Spend 45 minutes traveling three miles.
Carry your groceries home six blocks because parking near your apartment is more of a philosophical concept than an actual possibility.
Have “owning a washer and dryer” become one of your primary measures of generational wealth.
Walk past a mountain of trash bags on your way to a restaurant with a Michelin star.
Tell your friends back home you don’t need a car while spending $46 on an Uber because the subway is delayed again.
Go to Central Park when you want to remember what trees look like.
Spend your twenties telling everyone you could NEVER live anywhere else.
Then one day you visit your cousin in Iowa.
He has a house with four bedrooms, a garage, a yard, two cars, four kids, and somehow pays less than you do for the apartment where your refrigerator touches your couch.
But hey.
The NYC bagels are incredible.
dfi retweeted
Scaling self-verification with DeepSeek V4 Flash beats Claude Fable 5 on Terminal-Bench 2.1, while being 11x cheaper 💰
As open-source models become more capable, they can now generate large numbers of high-quality candidate solutions and verify their own outputs at very low cost.
For example, we find that sampling just 5 solutions with DeepSeek V4 Flash and ranking them using the same model with LLM-as-a-Verifier can lead to a significant boost in accuracy (79% → 88%), outperforming closed frontier models on Terminal-Bench.
Try it out today: github.com/llm-as-a-verifier…
More on verification scaling in my previous post.
How can we extract richer signals from AI Feedback?
Introducing LLM-as-a-Verifier✨— a simple verification scaling framework that achieves SOTA on agentic benchmarks 🚀
The key idea:
- Use fine-grained scoring granularity (e.g., 1-20 instead of the standard 1-5 scale)
- Take the expectation over the full logprob distribution of score tokens
- Scale repeated evaluation and criteria decomposition
You can use these fine-grained signals for more effective test-time scaling, RL, and agent monitoring! It achieves SOTA across Terminal-Bench V2, SWE-Bench Verified, RoboRewardBench, and MedAgentBench 👑
Advised by @Azaliamirh @istoica05 @drmapavone @chelseabfinn
🧵👇
dfi retweeted
During LLM inference on the DGX Spark (GB10), vLLM keeps 3–4 CPU cores at 100%. They aren't computing anything.
Changing one default (busy_loop_s: 1s → 2ms) cut vLLM's CPU from 333% to 89% and the SoC by 11°C. No measurable performance change.
nacyot.github.io/artifacts/v…
dfi retweeted
PSA:
this is verified and should be implemented:
"DGX Spark에서 LLM 추론 중 가장 뜨거운 건 GPU가 아니라 CPU였다. py-spy로 추적하니 vLLM의 대기 루프가 코어 4개를 최대 클럭으로 헛돌리고 있었다. 기본값 한 줄을 고쳐 CPU 낭비 12분의 1, SoC 최고 97→73°C. 성능 손실은 없다."
nacyot.github.io/artifacts/v…
