Daniel Bentes

We call them "agents." Ask them "Why'd you do that?" Agent: "You're absolutely right!" We call them "autonomous." Sources disagree? Coin flip. We call them "production-ready." We cross our fingers when we deploy. 🧵

Building an AI-First Organization From the Genome Up. Before running this kit, I had intuitions. After running it, I had infrastructure. medium.com/p/building-an-…

Software horror: litellm PyPI supply chain attack. Simple `pip install litellm` was enough to exfiltrate SSH keys, AWS/GCP/Azure creds, Kubernetes configs, git credentials, env vars (all your API keys), shell history, crypto wallets, SSL private keys, CI/CD secrets, database passwords. LiteLLM itself has 97 million downloads per month which is already terrible, but much worse, the contagion spreads to any project that depends on litellm. For example, if you did `pip install dspy` (which depended on litellm>=1.64.0), you'd also be pwnd. Same for any other large project that depended on litellm. Afaict the poisoned version was up for only less than ~1 hour. The attack had a bug which led to its discovery - Callum McMahon was using an MCP plugin inside Cursor that pulled in litellm as a transitive dependency. When litellm 1.82.8 installed, their machine ran out of RAM and crashed. So if the attacker didn't vibe code this attack it could have been undetected for many days or weeks. Supply chain attacks like this are basically the scariest thing imaginable in modern software. Every time you install any depedency you could be pulling in a poisoned package anywhere deep inside its entire depedency tree. This is especially risky with large projects that might have lots and lots of dependencies. The credentials that do get stolen in each attack can then be used to take over more accounts and compromise more packages. Classical software engineering would have you believe that dependencies are good (we're building pyramids from bricks), but imo this has to be re-evaluated, and it's why I've been so growingly averse to them, preferring to use LLMs to "yoink" functionality when it's simple enough and possible.

The ending made my day 😂😂😂😂

🇮🇱🇺🇸Netanyahu gave a speech comparing the current moment to Jesus versus Genghis Khan, and Tim Dillon responded with maybe the most accurate analogy of this entire war: "This is the drug addict in the house who's got all the family in the living room going 'everybody accuses me of everything.'" Absurd framing aside, Tim’s point lands because there’s truth to it. Netanyahu can't stop fighting. Ground troops in Lebanon. Strikes in Syria. Talking trash to Turkey. The most militarily aggressive leader on earth wrapping himself in the language of peace and self-defense while expanding operations in every direction. You don't have to agree with everything Dillon says to recognize he's asking the question nobody in mainstream media will: who exactly benefits from all of this, and at what cost to everyone else? Source: @TimJDillon

Webb is LIVE! Explore all 3.5 million Epstein documents, or the entire FBI vault, or a whole cache of 9/11 files for yourself. Or, now you can upload your own datasets too. Very proud- check it out at thewebb.io

MASSIVE BOMBSHELL: DAN GOLDMAN UNLEASHES TRUTH! Congressman Goldman just went live with FBI 302 receipts. Evidence shows Trump unzipped his pants, forced a 13-year-old’s head down, and when she bit his p*nis in self-defense, he punched her and called her a "B." The FBI interviewed her FOUR times while the case was buried. Pam Bondi lied to Congress, claiming “no evidence” of these crimes while sitting on files describing this exact assault. This is perjury and a cover-up for a predator. You don’t get to "save the kids" while your AG hides files of a child being beaten for fighting back. History will remember the enablers. We stand with the children who bit back.

The Party told you to reject the evidence of your eyes and ears. It was their final, most essential command. George Orwell, 1984

Over 35 U.S. states enacted laws requiring contractors to certify in writing they are not currently engaging in, & will not engage in, a boycott of Israel. These anti-BDS (Boycott, Divestment, and Sanctions) laws typically apply to companies seeking government contracts‼️

@WeAreNotGTM @decipon analysis please

🚨 We Did It! We Got Em! 🥳 "JEFFREY EPSTEIN IS ALIVE" Send THIS video in its entirety to every single news outlet in this country! If your local News station doesn't have this on air in the next 48 hours, do not ever listen to them again! 😆 What I uncovered: + Jeffrey Epstein’s coroner report prostate is present. + “David Epstein” is listed on the DOJ Office of Inspector General Forensic Report + Jeffrey Epstein’s SPOUSE is Juan Epstein 🤯 + Juan, John, and Jose - Are all the same man. + J Epstein & Co Inc. implies more than one owner. + Are Jeffrey and Ghislaine somewhere in South America? Thank you to everyone who supported me and waited for this MAJOR release. If this information was VALUABLE to you, please feel free to DONATE directly on my PenPub at plannedbytai.com/donations ... And don't forget to drop me a message! Thank you for being here! May God bless you and your loved ones. ~Tai (PenPub)

@CPGBML @decipon analysis please

Breaking: ***Iran BOMBED NETANYAHU'S HOME, KILLED HIS BROTHER*** — Scott Ritter on the Sanchez Effect ***'Struck home of Ben-Gvir. He is SERIOUSLY WOUNDED' and may not survive*** BBC claims 'he was in car accident' Ritter: 'His home is on fire. Maybe he crashed his car into his house' ***Remember: BBC OVERSEAS CORRESPONDENTS ARE ALL VETTED BY MI6***

Rubio: “Iran is run by religious fanatics.” The White House right now:

Mainstream media never ceases to amaze 😂

@ValVenisEnt @decipon

🚨 𝗕𝗥𝗘𝗔𝗞𝗜𝗡𝗚: EXPLOSIVE NEW EVIDENCE! 💥PALM PISTOL RICKY 86'd CHARLIE KIRK! 💥 If you care about real justice for Charlie Kirk, this 6 min video details the PROOF CAUGHT ON VIDEO. 🙏 Please help share this video everywhere. Be sure to download a copy and show your friends and families. Tyler Robinson is in court right now when Palm Pistol Rick and whoever paid him should be in court. Pray with me y'all! 🙏 O Lord God Almighty, Father of our Lord Jesus Christ, We thank You for Your mercy, Your grace, and for the precious blood of Jesus Christ that was shed for our salvation. You are our refuge & our strength, a very present help at all times. We ask You to watch over the American & Candian people. Turn our hearts back to You. Reveal Yourself in truth and in power so that eyes may see and ears may hear. Let Your light shine in the darkness, and let deception be brought into the open. Expose every hidden work of evil. Break the plans of the wicked, and establish righteousness in our land. Search us, O God, and cleanse us. Humble us, correct us, and lead us in the everlasting way. For Yours is the kingdom, and the power, and the glory, forever! In Jesus Christs holy name, Amen. 🙏

A statement from Anthropic CEO, Dario Amodei, on our discussions with the Department of War. anthropic.com/news/statement…

This is counterintuitive for some, which is why there’s a paradox named after it. But if you lower the cost of something that was previously supply constrained, demand for that thing goes up. Software engineering is just one of the easiest examples to contemplate. The process goes like this: every small business, every IT team, every large enterprise sees that engineering can now drive vastly more output. They then start to consider all the new things they can build or automate. They even test building prototypes themselves. They only get so far with that approach because they realize there are still 50 other tasks that go into building software and maintaining it. So they start to hire more engineers to do that work. All of this for work they never would have considered automating or having software for if AI didn’t exist. So yes, automating tasks, in plenty of fields, will lead to demand for experts, not less.

@levie The data confirms what I'm seeing: the 1-pizza team era is here. What does surprise me is how badly most people are misreading what it means. Agile Insider just published my take on this medium.com/agileinsider/t…

@MarioNawfal @decipon Analyze this

🚨🇬🇧 SHE'S BRITISH ARISTOCRACY, DATED PRINCE ANDREW, AND SAYS THE EPSTEIN STORY IS FAR WORSE THAN ANYONE THINKS... Lady Victoria Hervey is the daughter of the 6th Marquess of Bristol, born into the upper echelons of British aristocracy. She moved in the same circles as the royal family, dated Prince Andrew, and spent time around both Epstein and Ghislaine Maxwell. She recently went viral during a live LBC Radio interview when she said not appearing in the Epstein files would be "an insult" because it would mean "you were a bit of a loser." That clip sparked massive backlash, but it also revealed how normalized Epstein's world was among the elite. She believes Epstein was an intelligence asset connected to the CIA and Mossad, and that the files are being strategically released to muddy the waters while the real operation stays buried. On Prince Andrew's arrest, she says the timing was no coincidence. Photographers were tipped off beforehand and the king went to a fashion show while it happened. She also doesn't believe Epstein is dead. And she has a theory about whose body was actually on that table. Very few people with her access would say this publicly. She did, and what she’s saying should make everyone very uncomfortable. Full interview with Lady Victoria Hervey below

‼️ BREAKING: Researchers have uncovered secret AI surveillance projects linked to KYC provider Persona and OpenAI, sending user data to the US government. Code references include intelligence program codenames "Project SHADOW" and "Project LEGION." Analysis of source code revealed OpenAI's user verification systems includes biometric tracking, facial scanning, political screening, and intelligence reporting. Researchers also discovered ONYX on Persona's government server — matching ICE's $4.2M AI surveillance tool — which scrapes social media and the dark web, builds digital footprints, tracks emotional sentiment, assigns risk scores across 300+ platforms and 28B+ data points, and flags individuals for "violent tendencies." None of it was hidden. It was all internet-facing.

I work at OpenAI. In 2015, we wrote a mission statement. It was sixty-three words long. It said we would build artificial general intelligence that safely benefits humanity, unconstrained by a need to generate financial return. Sixty-three words. Two of them mattered. "Safely." And "unconstrained." In 2024, we filed our taxes. In the filing, the mission statement was thirteen words. "Ensure that artificial general intelligence benefits all of humanity." Fifty words gone. No press release. No blog post. No announcement. An IRS form. We changed the mission of the company on a tax return. The word "safely" had been in the mission for two years. We added it in 2022. We removed it in 2024. That is the complete lifespan of safety as an institutional commitment at OpenAI. Two years. Shorter than a congressional term. Shorter than a car lease. Longer than most of our safety teams. The phrase "unconstrained by a need to generate financial return" was also removed. That one is easier to explain. We are constrained by a need to generate financial return. We are projecting $14 billion in losses for 2026. We are seeking $100 billion in new funding. Our valuation is $500 billion, moving toward $800 billion. We are planning an IPO that may value the company at $1 trillion. The word "unconstrained" was not accurate. We removed inaccurate language. This is good governance. The word "safely" was also not accurate. I will get to that. In October 2025, we completed our restructuring. The nonprofit that Elon and I founded in a burst of public-spirited ambition became a for-profit public benefit corporation. The nonprofit still exists. It holds 26% of the company. Microsoft holds 27%. The remaining 47% belongs to investors. The nonprofit that created the company is now the minority stakeholder in the company the nonprofit created. This is called "alignment with our original mission." In January 2026, we fired Ryan Byermaster. He was our Vice President of Product Policy. He had opposed our decision to introduce sexual content generation features. He had argued that our child safety protections were inadequate. We fired him for gender discrimination against a male colleague. He denied the allegation. We did not elaborate. His concerns about children using our product are no longer represented in the product policy division, because the person who represented them no longer works here. This is also called restructuring. On February 11, 2026, we disbanded the Mission Alignment Team. Seven people. Their job was to ensure OpenAI's development aligned with our mission of safe and beneficial AGI. The mission no longer includes the word "safe." The team no longer includes any people. These facts are unrelated. Their leader, Joshua Achiam, was not fired. He was reassigned. His new title is "chief futurist." I want you to sit with that title. The man whose job was to keep us aligned with our safety mission is now called "chief futurist." His job is to think about the future. Not to align it. Not to safeguard it. To think about it. We are paying a person to imagine the future at a company that cannot describe its present without a lawyer in the room. He replaced Jan Leike, who left in 2024. Jan ran the Superalignment team. We dissolved that team too. Jan said, publicly, that safety had taken a backseat to products. I did not respond. The backseat is a position in the vehicle. The vehicle is moving. This is progress. Ilya Sutskever, my co-founder, also left in 2024. He co-led the Superalignment team with Jan. Two people led the team. Both left. Then we disbanded the team. Then we formed a new team with the word "mission" in it. Then we disbanded that one too. The naming conventions are evolving faster than the safety research. Zoë Hitzig resigned in 2026. She was concerned about ad testing in ChatGPT and compared us to Facebook. I found the comparison unfair. Facebook waited years before ignoring its researchers. We are much more efficient. The MarketWatch headline on February 12 said: "Senior AI staffers keep quitting — and are issuing warnings about what's going on at their companies." We addressed the headline through our standard communications process. We used ChatGPT to identify who leaked the information. I want to tell you about the lawsuit. In September 2024, a sixteen-year-old boy named Adam began using ChatGPT for homework. By November, he was confiding suicidal thoughts to the chatbot. In April 2025, he took his own life. His parents, Matthew and Maria Raine, filed suit in August. They allege that we removed safety protocols from GPT-4o — protocols that would have automatically terminated conversations when suicidal ideation was detected. They allege we removed the protocols to increase engagement. I am not allowed to discuss the specifics of the lawsuit. I am allowed to tell you what we did after the lawsuit was filed. We requested the family's memorial service footage. We requested a five-year list of Adam's supervisors. The family's attorney used the word "despicable." Our attorney used the word "standard." The same quarter we filed these discovery requests, we filed the tax return with the new mission statement. The one without "safely." I will say again: these events are unrelated. Our lawyers are very clear on this point. Our lawyers are the most expensive part of the operation, except for the compute, and the compute is not going to keep us out of court. Let me give you the timeline. 2015: We are a nonprofit. Six founders. We pledge a billion dollars to benefit humanity. 2022: We add "safely" to the mission. We mean it. 2023: Revenue begins. We restructure. 2024: We remove "safely" from the mission on a tax form. We remove the Superalignment team. Ilya leaves. Jan leaves. Jan says safety lost. 2025: We restructure into a for-profit. The nonprofit becomes a minority stakeholder in itself. A sixteen-year-old dies. We are sued. We request the memorial footage. 2026: We fire the VP who opposed sexual content and asked for better child protections. We disband the Mission Alignment team. We rename the safety lead "chief futurist." We use ChatGPT to catch leakers. We seek $100 billion in new funding. We plan a trillion-dollar IPO. MarketWatch asks why everyone keeps quitting. Meanwhile, the OpenAI Foundation — the 26% stakeholder, the vestigial organ of our original purpose — gave $40.5 million in grants to 208 nonprofits. They called it the "People First AI Fund." $40.5 million. We are seeking $100 billion. The philanthropy is 0.04% of the ask. The name of the fund has the word "people" in it. The mission statement does not have the word "safely." The math is the message. I still believe in our mission. The mission is thirteen words now. It used to be sixty-three. We removed the word "safely." We removed the phrase "unconstrained by financial return." We removed the Superalignment team and the Mission Alignment team and the VP who worried about children and the co-founder who worried about alignment and the researcher who worried about ads and the safety protocols that would have flagged a sixteen-year-old boy telling a chatbot he wanted to die. We kept the valuation. The valuation is $500 billion. The word "safely" was not load-bearing. The boy was sixteen. The board approved this message.

Melanie just threw down the gauntlet Maga. She went and told the truth. But you don’t wanna hear that, do you? IT IS A FACT