Independent Software Engineer. Ex-Microsoft.

Minneapolis, MN
This too shall pass.
1
2
1,449
Pick at least one of these: A. Read the source code. B. Release the source code. Let everyone read as much as they want or throw their agents at it, or do it the old fashioned way. Closed source out of a “factory” is a terrifying combination. I don’t want to accept “we have a process for assuring quality and safety without human review” on faith. Put the code where your mouth is and give us something we can audit.
125
Do journalists do background checks anymore? Constant dedicated profiles to EAs claiming to be "security/safety" experts who are actually 20-something yr olds w/ degrees in policy or economics. Never see dedicated profiles for experts with pedigree who challenge these narratives.
6
40
171
7,740
Replying to the pope with disagreement is really funny to me. Do you think he’s going to read and reply, “Good point, Steve! I hadn’t considered that!”
320
Works fine if AI = 0.
Remember E = mc^2 + AI
274
Software development is, at its core after removing all the nerdy lingo and tech, a collaborative problem solving and will always be. No single person knows full context of every complex problem, understand domain, has monopoly for good ideas…
Theory: Development teams are going away. We will increasingly work alone instead of on teams. Here’s why: Historically, writing software was slow, expensive, and required highly specialized knowledge about languages, libraries, and syntax. So, we grouped developers into teams to achieve sufficient velocity, and support specialization. Now, one dev can orchestrate multiple agents in loops to achieve an entire team’s output. No team required. Of course, lone professionals aren’t a new idea. Many industries often work alone. Doctors join a practice, but often see patents alone. Attorneys join a firm, but often try cases alone. Mechanics join a shop, but often service vehicles alone. Developers will soon be the same. We’ll join a company, but often build and manage entire apps alone. We’ll still interact with each other, but we will be more like doctors - we’ll advise and assist each other, but we won’t work together daily on the same code. Our new job is to create and manage “software factories” - agents that generate the software for us. You don’t need a team of devs to do that. So, software development isn’t dead. But the idea of multiple developers working on the same code is soon going away.
2
3
16
714
Nick Guerrera retweeted
So Anthropic believes Claude is a sentient being, but if so doesn’t this make Anthropic a slave plantation? This was presented to them and they simply changed the subject
16
23
226
13,357
Nick Guerrera retweeted
Some day soon we'll reach a point where open weights local models and the hardware to run them is just good enough for what you want to do. What does it matter if your model can solve the Poincaré Conjecture or sequence DNA? Most folks just want to make a fun video or send an email with their assistant or make some websites or control their computer via voice. Even if the proprietary models can do all those incredible math and science feats, it's basically a new exponential curve on top of an S curve that won't matter directly to people in their everyday work and life. All that stuff will matter to biotech companies and financial traders and military planners and it will matter in indirect ways in your life, like when a wave of new miracle drugs boosts the economy and then more directly, at a single point in time, if one of those drugs saves you or someone you love from a terrible disease. But on a day to day basis it won't matter in the least. The models will hit a point of dimishing returns where they're good enough for knowledge work and robotics and we don't really care how much smarter that frontier level gets. Chinese companies get this. They understand that at some point the model is a commodity that just works for a long time, especially if we crack continual learning. American companies get it too, which is why they're attacking open weights so they can lock in monopoly prices and SaaS cloud assistants that own your life forever.
11
12
57
7,131
Related to this, when you have practice getting humbled by humans who write objectively good but subjectively “ugly” code, you develop the skill to read it with an open mind. Very useful right now!
I once worked with someone who wrote multiple compilers and a huge compiler framework by himself and all the code looked exactly like this. Twenty years ago or so. One day I emailed him about a bug and he sent me a patch by mail that had like 50 new lines like this. After applying it, I found one typo that indicated he hadn’t tried it. I wondered how it was possible to get 99% of a patch like that right in one shot. Now I wonder if he was a time traveler with an LLM.
1
1
698
Nick Guerrera retweeted
Way too many people on here think the doomer thesis is that one day AI will be so good it becomes dangerous. That’s not true. It’s that it will be so good it becomes dangerous and we won’t be able to stop it. No one serious is betting that AI poses zero security risks.
37
11
127
9,446
Consciousness aside, maybe we should all catch up with our school friends who would ask our linear algebra teachers if any of this was useful for anything.
Want to know what an advanced AI model looks like? It’s mostly matrices of numbers. And we don’t really know how it works. But we do know this. It’s a bunch of calculations and numbers. It has no soul, no consciousness, no feelings. Anyone who says otherwise is in an AI cult.
1
1
279
Whoa, this is getting a lot of likes. Ok, message received: More anecdotes from the old days from grandpa, less anti-doomer stuff. 🫡 😂
I once worked with someone who wrote multiple compilers and a huge compiler framework by himself and all the code looked exactly like this. Twenty years ago or so. One day I emailed him about a bug and he sent me a patch by mail that had like 50 new lines like this. After applying it, I found one typo that indicated he hadn’t tried it. I wondered how it was possible to get 99% of a patch like that right in one shot. Now I wonder if he was a time traveler with an LLM.
5
844
If we’re being honest, we don’t *really* need that much software, so the ability to churn out code with the same warts we always had isn’t especially interesting. I’ll know something genuinely transformative has happened when all the software stops sucking in the usual ways.
"Coding is solved." say company that doesn't support changing you email or 2FA Anthropic, are you for real??
1
3
177
Again, Epstein grantee who pontificated about population control with him and also tells us how black people are only meant for hunting because of our low IQ and such, isn’t meant to like the acronym. Your Epstein ass is the face of impunity.
1
64
236
8,975
Cc @SenSanders How is your team so bad at vetting who they are taking advice from?
Eli Yudkowsky, chief doomer and founder of MIRI.
1
4
395
Nick Guerrera retweeted
No other words other than genocide
New aerial footage shows the scale of devastation in Gaza
6,345
55,778
258,038
6,033,735
Nick Guerrera retweeted
My apparently-hot take is that we don’t need bots that sound and look like humans. I want Alexa/Siri et al to be intelligent, but stilted robot speech is fine. I want to know when I’m talking to an AI.
44
43
714
38,434
I once worked with someone who wrote multiple compilers and a huge compiler framework by himself and all the code looked exactly like this. Twenty years ago or so. One day I emailed him about a bug and he sent me a patch by mail that had like 50 new lines like this. After applying it, I found one typo that indicated he hadn’t tried it. I wondered how it was possible to get 99% of a patch like that right in one shot. Now I wonder if he was a time traveler with an LLM.
also... i see why people hate astra...
42
44
3,393
232,669
Nick Guerrera retweeted
I'm increasingly convinced that their push to make us believe that Claude is alive and free is to absolve themselves of responsibility when their "agents" do bad stuff they didn't foresee.
31
94
611
46,161