Builder

Joined February 2023
The Paradoxical Commandments
106
239
19
1,330
80,762
You've probably seen the video. A young Frenchwoman in Paris, in tears, begging looters to stop. "Why are you doing this? That's not what we stand for." Some accounts are mocking her. "She voted for it." I won't. Because what you see on her face is not stupidity. It's a disease that has infected the entire West, and it has a name: suicidal empathy. The term comes from Gad Saad, an evolutionary behavioral scientist at Concordia University in Montreal. A man who fled the Lebanese civil war as a child, and who knows what it looks like when a society falls apart. His new book, Suicidal Empathy: Dying to Be Kind, is the follow-up to The Parasitic Mind. Here is the thesis, and why it concerns every one of us. Empathy is one of the best things about us. It evolved for a reason: to care for our children, our kin, our tribe. It was calibrated for a world where the people you felt for were the people around you, and where empathy always came paired with its counterweight: the instinct to protect your own and to recognize a threat. Suicidal empathy is what happens when that mechanism misfires. When it fires at the wrong targets, in the wrong amounts, with no brake. When we feel more for the criminal than for his victim. More for the stranger who breaks the rules than for the citizen who follows them. More for how compassionate we look than for what our compassion actually produces. Where does it come from? Four layers. First, biology. Empathy is a spotlight, as the psychologist Paul Bloom put it. It lights up the one face in front of you and leaves everything else in the dark. The crying migrant on the boat is visible. The thousands of future victims of a broken system are statistics. A spotlight is a terrible tool for running a country. Second, our own civilization. The West is the civilization that put the victim at the center of its morality. Christianity taught us to care for the stranger, the poor, the weak. It was one of the greatest moral revolutions in history. But cut off from its other half, justice, responsibility, the duty to protect your own, it turns into something else. Third, ideology. In The Parasitic Mind, Saad described idea pathogens, ideas that hijack our reasoning. Suicidal empathy is the emotional counterpart: an ideology that hijacks our compassion. It sorts the world into oppressors and victims, then demands unlimited empathy for whoever ranks as the victim, whatever they do. Fourth, status. In the age of social media, empathy has become a display. Posting "refugees welcome" costs nothing and earns applause. The bill goes to someone else: the working class, the suburbs, the women walking home at night. The results are everywhere. Criminals treated as victims of society. Sexual assaults by migrants met with lenient sentences. Self-defense branded as aggression. Borders treated as cruelty. Saad's conclusion is simple: a society dies when it would rather display infinite tolerance than use its survival instinct. Now the part I'll add myself, and that few people dare to say. On average, women score higher than men on measures of empathy. It's one of the most consistent findings in psychology, across cultures, and it makes perfect evolutionary sense: for most of human history, the survival of children depended on it. It's a strength. But it means an ideology that weaponizes empathy will hit women harder. And the data shows it: in country after country, young women have moved sharply to the left of young men, especially on immigration. This is where men have a responsibility. Not to mock. Not to lecture. Not to say "she deserved it." Empathy has always needed a counterweight, and that counterweight was the protective instinct: the one that draws lines, says no, and stands between the threat and the people it loves. Men abandoned that role. They were told it was toxic, and many believed it. So they stayed silent at the dinner table, at work, with their partners, sisters, daughters and friends. That has to end. Men have to have this conversation, honestly and with respect: compassion that cannot say no is not compassion. It's surrender. And the first to pay for it are women. Many women already understand this better than anyone. Some of the bravest voices in this fight are women, precisely because they are the first to feel the danger in the street. A civilization rests on a balance: compassion and strength, openness and borders, mercy and justice. When one side swallows the other, the foundations crack. What we are watching in Europe is what happens when empathy has no counterweight left. Look at that young woman again. She isn't the enemy. She's what this ideology made of a generation that was told kindness would be enough. Our job is to make sure she's the last generation that has to learn it this way.
She voted for it. She welcomed them. Faced with the encore, she begs them to stop. In her heart she knows they will not. Everything she believed. That they were French like her, fighting for the same cause. That the ones who warned her were full of hate. That she would be safe. It all falls apart in this moment. She is terrified of what comes next. In her heart she knows. Only remigration can save her.
3,704
11,264
1,193
49,360
3,462,037
Had a very fruitful session at the Ghana 🇬🇭 Grok Build today. Very inspiring. Thank you @YawAntwiOwusu @cursor_ai @SpaceXSI for the wonderful opportunity.
1
1
4
120
A hot take: if you can choose a fourth option when your coding agent offers three, and even adding its recommended one. it shows how well you understand what you're building.
1
19
Goodbye to LeetCode? Google just updated its interview process. AI is now a part of it. They have introduced a new round called Code Comprehension. The idea behind this round is to drop the candidates into a 200-500-line codebase with Gemini available for use. This round evaluates: - Can you read someone else's code - Are you able to debug independently before utilizing AI - Can you catch when Gemini is wrong
170
296
185
5,113
768,259
Replying to @xenocosmography
Getting a certificate doesn’t make you anything, except now you have a certificate 🤦‍♂️
691
274
75
5,560
137,850
psam retweeted
Replying to @kunchenguid
I agree. Closed-source vendors should rethink and rebuild their agents if they truly want them to belong to us (the customers). From first principles: what makes open source (OS) different from closed source (CS)? Several factors, but the key one is that in OS the user can run the whole application on their own hardware. If CS vendors want to truly serve us instead of owning everything, users should own some part of the agents. Since data is the gold mine in the SI (formerly AI) era, that part should be the memory layer. An agent is model + harness + memory. CS vendors should let us own at least the memory layer for now, so migrating between vendors is seamless. An agent is only relatively productive if it has more context about its user and preferences.
1
17
Does this mean all “.ai” domains owners will now have to purchase “.si” ??
How to buy .si domains? I can’t buy it from godaddy.
52
What once was the peak, is now just the beginning. Paramount and Warner Bros. shaped over a century of culture. By combining them, we aren't rewriting history — we're equipping these iconic studios with a more powerful engine. Together, we are Skydance: a creative-first home for bold, quality storytelling. We chose this name for a few important reasons. First and foremost, as we bring Paramount and Warner Bros. together, we wanted to preserve what has made each of these studios iconic. Both have distinct identities, extraordinary legacies and brands that have resonated with audiences for generations. We never wanted a new corporate identity to diminish, alter or overshadow either one. Instead, we wanted a name that would give the combined company an identity of its own while allowing Paramount and Warner Bros. — and all our extraordinary brands — to remain in the spotlight. We have big goals for Skydance, and we intend to pursue them with passion, imagination and a willingness to take smart risks. At the same time, we will honor what makes Paramount and Warner Bros. special — giving both studios the opportunity to grow, tell more great stories and bring those stories to even broader audiences around the world, powered by the scale and capabilities of Skydance. I couldn’t be more excited about what we’re going to build together.
4,094
4,485
1,505
34,371
10,281,108
Livestream for Starship Flight 14 has started, amazing! xitter.cf/i/broadcasts/1qxvvenMP…
9
13
2
122
10,056
Meta’s Muse is a cool app, but I don’t see it doing anything peculiar that agents like Grok bot, Hermes, or Open Claw are doing. I think the reason it has gained popularity is because they’ve been able to get it into consumers’ hands via a mobile app. If @NousResearch’s Hermes agent can also build a mobile app and distribute it to less technical people, they may have a pretty good shot at winning, because Hermes is currently 🔥
2
1
83
psam retweeted
Liftoff!
476
1,508
281
16,593
7,666,167
psam retweeted
this is literally steve jobs coded
mark zuckerberg might be the closest thing we have to steve jobs in this era
247
459
152
10,518
841,501
Being close to the customer is where the real profit is. I definitely have to cook something
A CS undergrad asked me where he could have most effect in the AI age. I said probably at either extreme: either close to the technology, actually making LLMs, or close to the customer, using AI to give them exactly what they want. Or maybe both if you can stretch that far.
1
42
A CS undergrad asked me where he could have most effect in the AI age. I said probably at either extreme: either close to the technology, actually making LLMs, or close to the customer, using AI to give them exactly what they want. Or maybe both if you can stretch that far.
411
951
126
11,475
965,786
Replying to @gyaigyimii
I don’t like how every new government is measured against the government that came immediately before it. So if NDC does better than NPP on something, we celebrate. When NPP does better than NDC on something, their supporters celebrate. Ghana is bigger than that comparison. NDC and NPP have alternated in power throughout the Fourth Republic. So after all these years, our question cannot always be “Are you better than the other party?” It should also be, are our lives getting better? Are the systems getting better? Are we solving problems permanently? There is nothing wrong with acknowledging progress. We should but praise should never replace accountability. If our standard is always the last government, Ghana only needs to improve a little. If our standard is the Ghana we know is possible, we keep demanding better until the country actually works. 69 years is a long time
18
72
15
384
66,666
Today I woke up to some of the best news I’ve ever had. Hermes from @NousResearch has been my daily driver for about 4 months. Today I officially contributed to the Hermes Agent repo. This means a lot. After months as a heavy user, I can finally say I’m also helping the project grow. Words can’t explain how happy I am. Everything great starts small — this is just the beginning of many contributions to Hermes Agent and the open-source community. I also want to thank you to @Kshitijjkapoor for the review of the PR.
Replying to @NousResearch
I’m proud to be a part of this community as a user. 💯 hopefully be a contributor soon
4
19
4,142
Am I the only dev who has no preferred display format (CLI/GUI) and just picks one based on mood and vibe?
Am I the only dev that prefers gui like cursor agents over cli ai tools like codex and claude code?
70
A little info to help you understand the "can't generate text" aspect of Jev. In the demo you see on the left side, the words and labels you see in the demo output (question names like “Revenue currently impacted?”, category values like “degraded” or “watch”, and fields such as “noul”, “choice”, “score”, and “confidence”) come from two sources you supply in advance: the question keys you write and the constrained answer space you define (a fixed list of options, a scoring rubric, or a yes/no probability). Jev only fills in the typed values and calibrated probabilities; it never invents new strings or explanations. Think of a normal LLM as a coworker who writes you a paragraph. You then have to read it, parse it, hope they didn’t invent a field, and turn it into something your code can use. Jev is more like a function you call from your backend. You give it: • some state (ticket text, user message, game state, logs, etc.) • a list of questions you already defined It gives back typed answers only: • yes/no probability • pick one option from a list you provided • a score on a scale you defined • plus a confidence number That is why the demo still shows words. Those words were not “written” by Jev. Jev just fills in the chosen value and a probability. Use it anywhere you currently write brittle if statements, regex, keyword matching, or “ask GPT and hope the JSON is valid.” Some best and worst case scenarios of using Jev Good fits: • route support tickets • flag risky actions before an agent does something • score emails/leads/PRs • decide whether to auto-reply or escalate • real-time decisions that need to finish in ~100ms Bad fits: • write emails • generate code • explain why • chat with a user A simple way to remember how Jev is different from LLMs: LLMs produce text for humans. Jev produces decisions for code.
Replying to @CompleteSkeptic
The gains aren’t free: Jev can't generate text Comparing Jev vs LLMs side-by-side makes the trade-off clear Fun fact: replacing sequential computation with parallel is the same way Transformers leapfrogged RNNs
106
Last month I wrote about how we can build a positive and safe future for everyone: meta.com/thefutureisforevery… Every lab has the responsibility and incentive to move at the pace required to train its models safely, and the ability to take its own actions to ensure that happens. The reality is: - People won't want to use agents that are misaligned with them and that don't do what they ask, so labs have a strong natural incentive to make their models more aligned. There is a lot of debate about slowing progress on capabilities until alignment catches up. My view is that trust and alignment are quickly becoming the most important capabilities that will differentiate agents and models. Any lab that doesn't focus on alignment will fall behind. - Labs face significant liability if their models cause harm, so they have a strong incentive to prevent this as well. Meta delayed shipping Muse for several months to focus on safety and security. We didn't call for everyone else to do this before we would. We just did it as part of our day-to-day work because it was clearly the right thing for people and for us. I'm proud of the security foundations we've built. - Engaging independent evaluators and advisors is industry best practice. MSL already does this today in several areas because it helps produce better work. Other labs can just do this too. In general, it would be helpful for there to be a larger and more diverse ecosystem of evaluators. - Committing the significant majority of compute towards serving people rather than racing towards recursive self-improvement is one of the best ways to ensure we develop this technology safely. Meta has made this commitment and other labs can do this as well. I believe the key to building a positive future for everyone is maintaining the right balance of power. This is within our power to do.
1,931
2,845
1,554
28,373
7,782,320