🧀 Jordan Love: 23-of-29, 372 yards, 3 TDs
🧀 Christian Watson: 7 catches, 171 receiving yards, TD
🧀 Matthew Golden: 10 catches, 164 receiving yards, TD
A ridiculous day for the @Packers offense ⭐️
@unsigned_sh0rt Personally I really enjoy reading your blogs and similar researchers, and the way you guys write is always more enjoyable than LLM-generated content.
Dusting off this tweet from April because this gap in shared understanding of LLM capability is *widening*. It's now less "two groups of people speaking past each other" and more a sharp funnel.
- Napkin math somewhere around 6B people (~75% of the population) have barely come in contact with LLMs at all.
- Around 1-2B (~20%) are casual and infrequent users of free-tier ChatGPT-like products. This group sees derpy chatbots and treats them a bit like a better Google search, a writing aid, or etc. Maybe an agent tries to book you a flight. I have non-tech friends who (reasonably, imo) say they have not much use for it at all. Even many professionals outside of math&code are in this tier. For example, execs and many other functions spend a lot of their time talking to other people, so while they understand what is happening intellectually, it is still second-hand and a bit abstract.
- Now we get to professional use of frontier-grade LLMs in math&code. Somewhere around 20M people (0.2%) see first-hand that large, complex projects that used to take them weeks/months can now be completed by agents with a prompt. Building apps, copying apps, translating apps, decompiling apps from binaries... This has all happened very quickly and recently - less than 1 year ago, I was writing code manually by hand, typing memorized computer code commands into a code editor character by character, occasionally pressing Tab to autocomplete a little chunk of code.
- And finally we get to the ~5,000 people (~0.00006%) with access to frontier-grade systems internally. The external world has seen the preview. It looks like swarms of thousands of agents collaborating over weeks on software mega projects: minting zero days, running cyber attacks and defenses at machine speeds, discovering new science, advancing the frontier of mathematics. Things that would have taken top professionals in the industry years of work. Meanwhile, human review and comprehension are starting to fall behind. For example, people are still involved in the "archeology" of the OpenAI-HF incident from many months ago. Mathematicians may be poring over the 722 manuscripts on frontier mathematics for a while.
The funnel is driven by a combination of factors:
1. The impact scales with ambition, problem size, and horizon. A question with a paragraph answer barely stresses the system. You need a reservoir of big, difficult problems that you really care about. This aspect drives the consumer / professional dimension of the funnel.
2. The jaggedness of the system (which I have written about a lot separately). Capability peaks in domains that are digital, verifiable and economically valuable. This is because LLM capability emerges from reinforcement learning on verifiable rewards on a curated environment mixture driven by revenue potential. This aspect primarily drives the area (e.g. math&code) dimension of the funnel.
3. Access. Free-tier, paid-tier, internal.
So this is the weirdness of the moment. The general public has mostly not interacted with these systems. When they have, it looks like a derpy chatbot. The majority of professionals still see only a modest uplift. And a small sliver of professionals are experiencing the vertigo of the curve going vertical. And it is all happening at the same time.
AI is cool and badass. I like it.
No, it's not the end of the world. We can "contain" it.
Now I will rant for a moment about what I dislike about AI.
I strongly dislike when large corporations try to inject AI into every product, essentially forcing us to use AI.
I also strongly dislike when people use AI for every small little thing in their life, such as leaving comments on websites like Xitter, YouTube, Reddit, etc. It defeats the entire purpose of leaving a comment on something... you're supposed to share YOUR thoughts, not an AI agents generated string output. It's goofy.
I also strongly dislike the nonsense blabbering of AI propaganda, how it will destroy the world, or how companies need to "contain it", essentially acting as means to corner the AI model market, trying to strip it away from individuals who may want to host it locally. It's a facade of "Only WE can do AI, not YOU", whilst charging absurd fees for usage.
I also dislike AI nerds who try to tell me to use AI to do things I do simply because I enjoy doing them. Yes, maybe an AI agent could reverse engineer faster than me, or code faster than me, but I do it because I enjoy it.
AI is an excellent tool to enhance work flows, or learn (assuming it doesn't hallucinate).
Okay, that's all I want to say about AI for today.
Here is my stance on Matt LaFleur:
He was hired in 2019 on a 4 year contract.
In his first 3 seasons his teams won 13 games every year, reaching the NFC Conference Championship twice and earning three 1st round playoff byes.
He was rewarded with a 4 year contract extension in 2022.
On his second contract his teams won 54% of regular season games. Missed the playoffs one year, was the 7th playoff seed three straight seasons with one total playoff victory.
This is a major decline.
2025 specifically showcased several epic collapses which unfortunately became expected as falling apart late became baked into the team identity.
I fully thought it was time to move on from Matt LaFleur after the 2025 season. The team performance since the first extension was not good enough.
I did not need to answer “who would you hire instead?” to know I didn’t think he was the right guy to lead this team. It’s ok to dump somebody you like if you fell they are not “the one” for you long term without knowing who you will date next on your search for “the one”.
The Packers instead pointed at his entire Packers resume, highlighting his early success, in their decision to extend him a second time.
To me this the main issue. His early success earned him his first extension - rightfully so.
But the measuring stick, in my mind, should be the performance during the extension years. His initial years already “did their job” and should be background consideration instead of once again being the focal argument for extension.
Now, when Matt LaFleur was hired only 6 teams per conference made the playoffs each season. On this same standard the Packers would have missed the playoffs in each of his seasons on his second contract.
Some don’t like this 6 vs 7 teams comparison. Ok, fine.
For me, if you want to look backwards to his early resume success as a measuring stick, then we should apply the same measuring stick to his teams’ recent performance - which would have been 4 straight non-playoff worthy performances every year of the extension period, apples to apples.
If you don’t want this comparison, then let’s focus entirely on the second contract performance - which again led to 1 total post season win in 4 years - and set aside the early resume.
So to me, using either thought process, it just wasn’t good enough and I didn’t see any reason to expect improvement given the cultural collapses on the field which seem baked deep into the Packers DNA.
The Packers, however, decided to reward Matt LaFleur to a second contract extension with an uncertain number of more years, making an uncertain, yet significantly higher, amount more fully guaranteed money.
So far, nothing I have seen this 2026 has suggested the Packers made the correct decision to re-invest into Matt LaFleur.
Jameis Winston on being Giants starter:
“I gathered the guys and said, if one man has one crab leg he’s a mere mortal. But if that same man has 53 crab legs he’s an unstoppable force.”
@h4x0r_dz Highly doubt this is an unauth. PetitPotam and all of the RPC interfaces for coercion have required auth since 2021. See my open issue on the repo which the author has not answered
@_xDeJesus re: CVE-2026-62911 - I think the unauth claim is wrong. I opened an issue on the github repo, but the VRED writeup demos it requiring a lowpriv user. PetitPotam uses the EFSRPC interfaces which have required low-priv auth since 2021