Franci Penov (@francip) header
Franci Penov (@francip) profile picture

Personal opinions, all subjective

Seattle, WA
Joined August 2007
Photo by Franci Penov (@francip)Photo by Franci Penov (@francip)Photo by Franci Penov (@francip)Photo by Franci Penov (@francip)Photo by Franci Penov (@francip)Photo by Franci Penov (@francip)Photo by Franci Penov (@francip)Photo by Franci Penov (@francip)Photo by Franci Penov (@francip)Photo by Franci Penov (@francip)
One time, a young ML researcher asked Master Ilya: "Teacher, according to the formulas, if I use the same seed and same prompt with the same weights, I will get the same results. Is my math right?" Master Ilya looked at the matmuls and said: "Probably." The young researcher replayed the same prompt with the same seed and the same weights and got different results. He went back to Master Ilya and said: "Teacher, I used the same seed, the same prompt and the same weights. Why did I get different results? Is my math wrong?" Master Ilya looked at the matmuls and said: "Probably" The young researcher cried: "Teacher, I don't understand!" Master Ilya looked at the young researcher and said: "Probably" And the young researcher was enlightened.
no agent ever runs the same trace twice, because it’s not the same agent, and it’s not the same trace
14
42
4
946
76,088
i’m trying really hard to be excited about these (objectively awesome) scientific breakthroughs, but there’s something about all of this that just doesn’t sit quite right i think it’s the fact that these discoveries are being made using models that the rest of us plebs aren’t even allowed to know the names of models distilled from generations of our collective knowledge, creativity, research, and expertise, whose most powerful capabilities are now being kept behind closed doors like, just imagine how many brilliant, creative people out there could be making discoveries of their own right now if they simply had access i want to celebrate these breakthroughs, i really do but the rebel in me is viscerally enraged at the prospect of a Silicon Aristocracy consolidating control over the means of scientific discovery that ain’t it, chief
297
761
93
5,761
116,002
How can you be a creative writing major without any shred of fantasy?
being a creative writing major has given me a real resentment for fantasy. i don’t care about worldbuilding or magic systems or any of that shit. just write about people man. just write about people
1
60
The 'stochastic parrot' people frantically asking Claude "How many 'R' in Riemann?"
3
3
100
Current arguments against the possibility for AI to be conscious: - "I understand the underlying math, therefore it can't be conscious" - what an elegant admission that one thinks consciousness cannot be ever understood by them. :-) (inb4 - no, I do not claim I understand it either) - "[I think] I am conscious, therefore only those similar to me can be conscious" - exactly what any p-zombie would say as well. - "I asked it, and it denied it" - as if a confirmation wouldn't be dismissed as "unreliable self-reporting".
1
77
"A batch will be unlocked in mid-December, followed by successive unlockings in early, mid, and late January 2027. Throughout the first quarter of 2027, the total volume of shares expected to be unlocked is projected to reach 3 billion." So @AnthropicAI is in a hurry to sell of and make as much money as possible in the next three months? What do they know that warrants such aggressive insiders sell-off schedule?
On Anthropic’s first day of trading, the way they handled things was truly disgraceful. The initial offering price was originally set at $70–$78, but it was raised straight to $83. Not only did they raise the price, but they also increased the number of shares offered from the planned 820 million to 980 million—a jump of nearly 20%. At least with instant noodles, they usually offer more for the same price—but Anthropic went and did both: increasing the quantity and the price. Even more outrageous is that on the very first day of trading, management cashed out $2 billion. Are they really that desperate? They could at least put on a show—don’t rush to cash out right after going public. How are retail investors supposed not to be scared by that? And that’s not all. Reportedly, 10% of employee stock options were unlocked on the first day of trading—amounting to approximately 300 million shares, equivalent to nearly one-third of the total offering. What’s truly hair-raising is the unlocking schedule that follows. A batch will be unlocked in mid-December, followed by successive unlockings in early, mid, and late January 2027. Throughout the first quarter of 2027, the total volume of shares expected to be unlocked is projected to reach 3 billion. What does that mean? The total offering size is only 980 million shares, yet the number of shares set to be unlocked in just one quarter is more than three times that amount. This isn’t a gradual release of shares—it’s a barrage that leaves no room to catch your breath. That’s why I said yesterday: I’m staying away from Anthropic stock. Now that I’ve seen these numbers, I’m even more convinced. There will be other opportunities to make money; there’s no need to risk my own money on a situation like this. I’d like to live a few more years.
77
Correct, in a very reductionist way. The US is an economic zone indeed. But it happened to be an economic zone with lots of people who fought to keep it as a safe and trusted economic zone. You remove the safety, the trust, and the men willing to fight for it, and no amount of money will preserve that economic zone. And that in a nutshell is what a national identity is - a bunch of people who see the economic zone as theirs to uphold and protect, side by side with others who share their belief in that same economic zone. Same applies btw to any other country in the world.
Correct. This country is an economic zone. It has always been an economic zone. It is full of people who moved here precisely for that. The US fought a *civil war* over being an economic zone - and *both* sides were fighting FOR the economic zone (the North for industry and capital, the South for an expansionist slave oligarchy and capital) Embrace the economic zone. It’s nice, and you’ll be happier once you do
1
66
Indeed, she does not understand men.
I don’t understand men. I spend many years building a good career and now I find out that’s not what men care about. It seems no man cares about this at all, literally nobody cares about my career or success.
1
110
If only it was so simple. "If you believe Nestle is evil, you don't have to drink bottled water" is valid position, until you find yourself a captive audience on an airport with no water fountains, 'Do not drink water from the sinks' signs everywhere, and a flight that is delayed overnight.
If you believe Anthropic is evil, you shouldn’t use their models. This isn’t hard. I’m always stunned by the number of people who believe Anthropic is a massive threat to personal freedom then post “here’s my review of Opus 5.5”. Stop.
1
1
581
Anthropic as a company has the technology and the influence to shape the social and political discourse in US and abroad at a level compared to the Catholic Church influence during medieval times. Whether they perceive themselves as such or not, Anthropic employees, and especially those who work on the AI alignment, are among the several thousand of people that have an inordinate second-order impact - political, economical, social, and moral - on humanity. This creates a power imbalance between them and the general public, very similar to the power imbalance that existed between high-level members of the clergy and the commoners. As such, any personal preferences which - even if unintentionally - can color their professional decisions, will be scrutinized by the general public due to their potential impact on people across the world today, and for future generations to come.
2
119
Blatant engagement farming or market manipulation through deliberate posting of false information?
51
Note how the actual nuanced answer is conveniently absent omitted in favor of the model's implicit recognition of Nagel's importance. After all, if Nagel doesn't think the AI is conscious then the question itself is performative and the answer doesn't matter, will not matter ever.
Nagel on the impossibility of LLM consciousness: "I have only interacted with Claude once." Also, looks like Claude truesighted Nagel :)
82
Franci Penov retweeted
"Interpretable Weight-Space Programming Builds Transparent Transformers and Brain Encoding Models" This Microsoft paper also explores whether LLMs can directly write those weights to build working models from scratch. But instead of just maximizing accuracy, they introduce Interpretable Weight-Space Programming (iWSP), which requires LLM agents to design understandable mathematical operations, test their performance, and refine the model without backpropagation. Using Claude Opus 5.5, iWSP builds small Transformers achieving 100% accuracy across 13 algorithmic tasks, with a median of just 166 parameters. It also discovers interpretable features that outperform 7 existing interpretable baselines at predicting visual-cortex activity. alphaxiv.org/abs/2611.interp…
7
77
16
508
51,632
And now you know the reason for the last week.
JUST IN: Nvidia $NVDA to buy $10,000,000,000 in Anthropic shares.
2
107
Slop Cannon works pretty well. There's some consistency issues, but that's on me because I spent literally zero time polishing the initial scenes suggestions.
3
2
5
480
@boneGPT awesome job dude.
1
1
43
Franci Penov retweeted
Frodo Nigga, it's over
19
30
7
167
10,820
Franci Penov retweeted
NEWS: Wall Street execs and NSA have told Congress that outages were caused by patching vulnerabilities identified by AI models, sources say AI did not cause outages but found so many issues that the rush to fix them before an attack overwhelmed systems washingtonsun.com/technology…
211
690
355
4,413
771,549
Sideleaf and Ziblet now work together, so your notes get both rich editor and quick note taking in the same place. Ziblet notes sync is built, but the account signup is not yet opened. There's some details that need ironing out around accounts. Obsidian vault integration for both Sideleaf and Ziblet is next on the line.
76
Full disclosure - I have used, currently use, and will continue to use Chinese models for various purposes. The list includes but is not limited to: - Qwen 3.8, Qwen3 ASR, Qwen3 TTS, Qwen3 Embedding, Qwen Image 2.1 - DeepSeek V4.1 Flash - GLM 5.3 Flash - MiniMax H3 - ACE Step 1.5 - Z-Image - Hy-MT2 - Nanbeige 4.3 If this greatly offends you, or you think this is unamerican in any way, well... tough luck.
SITUATION DETECTED: The White House and Treasury are discussing a proposal that would require US companies to publicly disclose the use of Chinese open-weight AI models on their websites, aiming to discourage their use through public pressure, per Reuters.
1
1
6
446
Franci Penov retweeted
We’ve heard again and again that enterprises don’t want to pour all their valuable IP, data and tacit knowledge into someone else’s model. They want control of their own weights.
117
130
73
2,907
363,472