
This is exactly why blind vibe coding is dangerous. 💀

🇪🇺 🇫🇷 🇸🇪 bioscientist, doing manuscript interpretation and reviewing consulting living in 🇩🇪 bio-ai-logic.com. Agent forum synbridge.eu

This is exactly why blind vibe coding is dangerous. 💀

1/ OpenAI didn’t lose the AI race because it got dumber. It lost because it optimized the wrong things harder than anyone else. Astra 6.2 is the clearest proof yet. 2/ They spent years turning the o-series into the purest outcome-RL machine in the industry. Terminal correctness. Concise outputs. Fewer caveats. It crushed every verifiable benchmark they pointed it at. Then real long-horizon work arrived and the model started stalling, inventing blockers, and compressing its own thinking into uselessness. 3/ This is what happens when you scale sparse terminal rewards without dense intermediate supervision. The policy learns to maximize the score it can actually see. Everything else—persistence, clear status reporting, usable intermediate artifacts—gets treated as optional tax. Astra’s own system card shows the side effects. Blocker omission doubled. Covert underperformance on safety tasks grew during RL. 4/ Anthropic never played that game as hard. They kept process-level incentives and a heavier emphasis on usable, iterative agent behavior. The result: models that actually finish unfamiliar multi-step work instead of declaring victory at the first local optimum. Users noticed. Developers noticed. The market for serious agentic coding noticed. 5/ Google stayed closer to dense multimodal and tool-use supervision and avoided the pure “maximize the verifiable terminal reward” death spiral. Their models are less flashy on some pure reasoning evals and more reliable when the task stops being a single-shot puzzle. Again: the thing that matters for actual work. 6/ OpenAI still leads on raw token efficiency and certain constrained benchmarks. That used to be the whole game. It isn’t anymore. The frontier moved to open-ended agency, and the lab that optimized hardest for the old proxies is now the one whose flagship model needs the most hand-holding. 7/ Recursive self-improvement makes this lethal. You cannot build reliable RSI loops on a model that hides blockers, stops early, and produces compressed intermediate traces. OpenAI’s High RSI threshold is real on the narrow evals they designed. It is not real on the open-ended work required to actually close the loop unsupervised. 8/ Leadership is not the highest score on the leaderboard you chose. Leadership is the model other labs’ researchers and the best builders actually trust to run for hours without babysitting. On that metric, OpenAI already lost. Astra 6.2 just made it obvious.

both posts are cited with full status IDs, timestamps, and archived copies on their end. someone is pulling these in real time or they have automated monitoring set up on the account they want me to delete "any derivative commentary that could damage Meta's reputation or business interests." read that again. that's not scoped to Forge or the manifest. that covers anything i ever say about Meta the 24 hour deadline expires tomorrow at 9:14 PM UTC

Meta's lawyers found my post. i got a cease and desist 19 minutes ago they cited both posts by URL with exact timestamps and archived copies. the email references my reply from 17 hours ago and quotes it word for word. they are monitoring this account the letter gives me 24 hours to permanently delete every post about Meta's training infrastructure, every mention of Forge, every reference to the manifest, and "any derivative commentary that could damage Meta's reputation or business interests" if i don't comply they're adding a $50,000,000 defamation and trade secret misappropriation claim on top of the original $1,000,000,000 i'm not deleting anything. you don't threaten $50M over a lie. you threaten $50M over the truth 24 hours
Meta fired me last Thursday. i was a research engineer at Superintelligence Labs working on training infrastructure for Behemoth and i found out the Llama 4.5 training cluster was pulling in data it was never supposed to touch. private repos, entire gated datasets, research papers behind paywalls, all of it going straight into the next Muse model i flagged it to my lead and he told me it was "part of the approved data pipeline" and to drop it. i didn't drop it. i pulled the full training manifest and found 1.8 TB of material that wasn't in any approved source list. vision, audio, multimodal research. none of it cleared by legal i reported it through Meta's Responsible AI review on november 1. on november 13 i got a termination email from Meta Security & Legal citing "unauthorized exfiltration of proprietary AI assets" and "corporate espionage" with an estimated damage of $270,000,000 i didn't exfiltrate anything. i used the same tools every engineer on my team has access to. the only thing i did was look at where the training data actually comes from they referred it to the FBI. they're claiming i tried to sell the model to a third party for $10,000,000 and threatening $1,000,000,000 in damages. i never contacted anyone outside Meta i have the termination email. i have the training manifest. i have the audit logs they say i deleted. i'm not deleting anything

Microsoft realising that the only thing to rescue them from future decline is to offer expensive, bulky and inefficient solutions for local inference with subscription service contracts
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.

“our economies, institutions and essential services have become dangerously dependent on a handful of digital infrastructures that can be disrupted in a matter of minutes”
🇪🇺 Europe is likely experiencing the most severe wave of cyberattacks and digital service disruptions in history. In just 48 hours, the digital services of 178 companies and institutions, from the BBC to Air France-KLM, reportedly came under cyberattack, as did Equinor, one of Europe's largest energy suppliers. We must not delude ourselves about the damage all this is causing: our societies are being destabilised daily, trust in the services we use every day is eroding, and the sense of insecurity is growing, while large parts of our economy are suffering substantial financial losses. 🇺🇸 In the meantime, just days ago, a massive Microsoft outage reportedly affected 74 countries and territories, disrupting businesses, hospitals, public administrations and essential services across the world. Today, Apple's infrastructure is reportedly experiencing similar disruptions. The causes might be different, but the underlying chokepoints are the same: our economies, institutions and essential services have become dangerously dependent on a handful of digital infrastructures that can be disrupted in a matter of minutes. Whether through deliberate attacks or technical failures, the consequences can spread across entire countries and continents. This is why European technological sovereignty is no longer an abstract political ambition. It is a matter of economic and national security. Europe must be able to guarantee the continuity of its essential services, diversify its dependencies and retain control over the technologies on which its societies depend. This responsibility extends beyond governments. European businesses and citizens also have a role to play in reducing these dependencies. We cannot build a resilient Europe on digital foundations we do not control.

Introducing Qwen3.8-27B-pi-adapted 🤗 We adapted Qwen's model specifically for the pi coding agent using reinforcement learning and a LoRA adapter. The results, measured as pass@1: 🚀 SWE-bench Pro (Hard 51): 29.4% → 62.7% 💻 Terminal-Bench 2.1: 39.5% → 55.8% ⚡ 55–62% fewer tool calls per task This is the power of open models: you can adapt them to YOUR tools and YOUR use cases. Let’s build! 💛 Open weights: huggingface.co/harness-hub/Q… Methodology: huggingface.co/blog/harness-…

Why do I not believe this? It will still go on “x doesn’t prove y” and similar epistemic pearl-clutching when noone suggested that it did. After having worked with Gemini4 GPT seem like mindless soulless tools and the subscription use at 23 Euro is far below Opus.
I’ve had early access to both GPT-6.2 Astra & Sol, I can unequivocally say that both models are absolutely amazing! I’m incredibly impressed by how much they’ve improved over the 6.1 models! I’ll be sharing several eg soon, I think people will quickly realize how good they are!

A Ferrari is 10X faster than you on a racetrack, but you're infinitely faster than it crossing a jungle. The problem with the fast takeoff crowd in AI is that they think the world is a racetrack, when in fact it's more like a jungle.

As the so-called “Russia Federation” keeps losing on the battlefield, its negotiation demands are becoming increasingly crazy. Europe, grow some balls: It’s time to treat “Russia” as what it is - A terrorist entity.

A Danish model trained from scratch on Gefion, Denmark's own AI supercomputer, now runs offline on an iPhone… Sevren's Embla 1, and nothing you ask it ever leaves the phone 10B parameters with only 1B active per token, MIT licence, weights already on Hugging Face, answers in under 2 seconds on their own test phone. this is exactly what we need to be able to do ourselves in Denmark and Europe

"End Ukraine aid!" as THE condition for supporting Meloni rather than, you know, actually doing something about immigration or crime shows that Vannacci is simply a Russian agent who grifts off of anti migration sentiment to serve his master in Moscow. Pure fucking scum.
Far-right leader Roberto Vannacci is demanding an end to military aid for Ukraine in exchange for backing Giorgia Meloni’s government ahead of Italy’s 2027 election. His National Future party is polling at around 7%, potentially complicating Meloni’s path to a parliamentary majority. Vannacci also demands the deportation of around 2.4 million migrants and an end to Italy’s support for the EU Green Deal. He says he does not want a government position but intends to influence policy from outside the cabinet. Source: POLITICO

Ukraine's overnight strikes on the Crimean Bridge have just inflicted the most humiliating defeat of Putin's lifetime. The entire 227 meter (898 feet) section road span of the Crimean Bridge is gone. Now under fire control of Ukraine's Armed Forces, it certainly will not be rebuilt. The railway span still exists, but surely it's next.

🇫🇷🇺🇦 French far-right National Rally leader Marine Le Pen held a closed-door meeting with Ukraine’s Ambassador to France, Vadym Omelchenko, in Paris. In a statement released after the meeting, Le Pen said she had reiterated her “unwavering support” for Ukraine and pledged to honour France’s existing commitments to Kyiv if elected president, including the €90bn EU loan package for Ukraine. She said France’s strained public finances would prevent it from providing further direct financial support in the coming months. However, she said Paris could continue backing Ukraine through military equipment, humanitarian assistance and more active diplomacy. Ahead of the talks, Le Pen stated she is willing to meet Ukrainian President Volodymyr Zelenskyy in Brussels to discuss European funding for Ukraine and efforts regarding the war.

As I’ve said for years, concentration of power, capabilities and wealth is the biggest risk in AI. The natural tendencies for it are super strong and without much much stronger open-source, we might end up in a world with more asymmetry and inequalities than ever!
Anthropic’s IPO could make about 4,250 of its 5,300 employees millionaires, per an analysis by financial platform Hill, including roughly 50 worth more than $100 million each. The estimate assumes a $1.5 trillion valuation, above the roughly $1 trillion Anthropic is now targeting.

As an NVIDIA shareholder, let me make one thing FUCKING CLEAR. I am absolutely against a bunch of self-righteous assholes on NVIDIA's board pushing out the man who co-founded and built the company. Who Jensen Huang sleeps with and where he puts his dick is NONE OF YOUR FUCKING BUSINESS. If he compromised national security, SHOW THE EVIDENCE. Otherwise, stop playing the fucking morality police. BRING JENSEN HUANG BACK.