Ovidiu Ursachi 🇪🇺 (@ursachi) header
Ovidiu Ursachi 🇪🇺 (@ursachi) profile picture

founder of @yourmasernet - cybersecurity & NLP, views on politics, economy, or other topics

Europe
Joined April 2009
When Putin seized Crimea, both Europe and America were quite openly making calls, holding meetings, and sending signals to the Ukrainians who were in power at the time that there should be no escalation – let's just keep things friendly and leave things as they are for now. And we lost Crimea. After that, we lost part of Donbas. But the shooting in the east continued. There was no peace. People think we were always helped. That is not true. Help started coming after the full-scale invasion, but that was after eight years of Russia occupying our territory. That is a long time to figure things out and understand that Putin has no intention of giving anything back to us. At one point, some Europeans asked me: “Why do you think that if you withdraw from Donbas, he will go further?” And I answered: “Because at the beginning of the war, he wasn't seizing Donbas. He came to the capital. He seized parts of Zaporizhzhia and Kherson regions, Kherson itself, and Kharkiv region.” We have retaken a lot of territory, but he had seized different parts of our country. And now they want us to give them something because they are unable to seize it. We may have a smaller territory, less money, and fewer weapons, but we still have more dignity than many others. We will do everything we can to end this war, but it must end in a way that preserves our dignity. From an interview with ZDF. (2/3)
743
6,413
217
33,016
1,368,865
Ovidiu Ursachi 🇪🇺 retweeted
Polonia demonstrează diferența dintre o societate care își repară statul de drept și o țară ca România, unde capturarea justiției a fost lăsată să devină literă de lege. În timp ce Tusk și noua coaliție de la Varșovia curăță activ instituțiile de judecătorii ilegali instalați de PiS (dublerzy) și restabilesc legalitatea constituțională, la noi măcelul început în epoca Dragnea–Iordache a rămas definitiv bătut în cuie. ​Ce a făcut PSD-ul prin Comisia Iordache și „reformele” toxice din 2017–2019, lăsate nesancționate până azi: 1. Eliminarea raportului dintre instanțele românești și deciziile Curții de Justiție a UE (CJUE): Au introdus sancțiuni disciplinare aspre pentru judecătorii care îndrăzneau să aplice dreptul european cu prioritate în fața deciziilor CCR politizate. Practic, magistrații erau amenințați cu excluderea din magistratură dacă respectau tratatele UE. 2. ​Completurile de 5 judecători și cele de 3 judecători de la ÎCCJ: Au forțat prin decizii de CCR rejudecarea tuturor proceselor mari de corupție pe motiv că acele completuri nu fuseseră trase la sorți „cum trebuie” sau nu erau „specializate pe corupție”. Rezultatul? Ani de judecată au fost aruncați direct la coș, procesele s-au reluat de la zero și au intrat direct în capcana prescripției. ​ 3. Răspunderea materială a magistraților (Arma fricii): Au modificat legea statutului judecătorilor și procurorilor astfel încât aceștia să poată fi acționați în judecată de Ministerul Finanțelor și puși să plătească din propriul buzunar dacă o decizie de-a lor era infirmată ulterior. Niciun magistrat nu mai avea curaj să aresteze un politician de teamă că își pierde casa. 4. ​Pensionările masive și anticipate la 20 de ani vechime: Au creat o schemă legislativă prin care sute de magistrați cu experiență s-au pensionat la 45–48 de ani pe pensii speciale uriașe. Instanțele și parchetele au fost golite deliberat, rămânând mii de dosare pe capul a câțiva tineri fără experiență, copleșiți de volumul de muncă. ​ 5. Recursul compensatoriu: Sub pretextul condițiilor grele din penitenciare cerute de CEDO, au eliberat înainte de termen peste 20.000 de deținuți, printre care criminali, violatori și hoți recidiviști, care au comis rapid noi atrocități pe străzi. ​Anihilarea probelor SRI / decizia privind protocoalele: Au eliminat retroactiv toate interceptările tehnice realizate de SRI pe mandate de siguranță națională sau autorizate de judecători, chiar dacă înregistrările prindeau politicieni primind valize cu bani pe sub birou. Dosare grele de zeci de milioane de euro au rămas peste noapte fără nicio probă validă. 6. Invenția Secției Speciale (SIIJ): O armă de intimidare creată exclusiv pentru a lua dosarele de mare corupție din brațele DNA și a le clasa, permițând oricărui politician anchetat să facă o plângere fictivă împotriva procurorului său pentru a-i bloca ancheta. ​ 7. Măcelărirea Codurilor Penale: Încercarea constantă de ciuntire a abuzului în serviciu și interzicerea probelor esențiale (înregistrări video-audio, camere de supraveghere), menite să lase procurorii fără instrumente de investigație. 8. ​Transformarea Inspecției Judiciare în bâtă politică: Numirea conducerii prin OUG direct de la Guvern, folosită pentru hărțuirea disciplinară a fiecărui magistrat care a îndrăznit să protesteze pe treptele tribunalelor. ​ 9. Fereastra de prescripție: Pasivitatea calculată a Parlamentului, care a refuzat timp de 4 ani (2018–2022) să pună Codul Penal în acord cu deciziile CCR. Prin această complicitate tacită, mii de dosare de corupție, evaziune și delapidare cu prejudicii de miliarde de euro s-au închis definitiv, iar infractorii de calibru au scăpat curați. ​10. Blocarea promovărilor la parchetele cheie: Creșterea bruscă a vechimii necesare pentru DNA și DIICOT, lăsând parchetele fără personal ca să nu mai aibă cine trimite marii corupți în judecată.
24
74
5
422
8,991
Ovidiu Ursachi 🇪🇺 retweeted
A lot of people seem shocked by China's heavy-handed retaliatory actions on Mbappé. They assumed once the Chinese understood the context - that the Mao meme was self-deprecating, that the "Chinese dictator" line was not malicious - the whole thing would get walked back as a giant misunderstanding. Let me explain why this kind of thinking is wrong, and why they were never going to backpedal. What happened to Mbappé is the modern version of that famous 1937 Stalin photo with Nikolai Yezhov. After Yezhov fell out of favor, the photo got retouched so comprehensively that the space where he had once stood was filled in with canal water and railing. It was as if he was never there. It's the same playbook here - once the system decided someone or something or some event is a problem, the record gets airbrushed until the problem no longer exists. Dan Wang, in his excellent book Breakneck, describes China as an engineering state run by technocrats who excel at construction and large-scale execution. He contrasts this with America’s lawyerly society that is process-heavy, rights-oriented, and tends to get stymied by obstruction. The same mindset that builds high-speed rail, factories, and manufacturing process knowledge also extends into social engineering that treats society, narratives, and public sentiment as systems to optimize. Wang’s examples include the One Child Policy and zero-COVID, both of which locked in once launched because the system prioritizes outcomes and consistency over flexible correction. You'll notice that the system tends to lurch to the extremes. These numbers - One Child, Zero Covid - are seriously autistic in its drive and focus to achieve overtly neat outcomes that are humanly impossible. The targets are unrealistic and sound like engineering specs, because leaders that are literal-minded end up seeing human activity as flows that can be turned on or off with valves. Mao remains the founding leader in official historiography; jokes or “dictator” labels challenge Party legitimacy and the narrative of national rejuvenation. Back in 2018, the Law on the Protection of Heroes and Martyrs came into effect which criminalizes the slander of those officially designated as heroes and martyrs. So once the reaction to Mbappé's clip ignited (Weibo outrage, coordinated brand statements), however erroneous it was, however much of a misunderstsanding it was, the engineering state executed a full response that left no room for nuance or de-escalation: > Chinese state media canceled La Liga broadcasts indefinitely > Companies dropped him > His face was scrubbed from all ads in China > His products were removed from shelves > Mbappé has been removed from Chinese versions of major football video games > Platforms muted or restricted accounts and promotional material that mentions him. While a lawyerly society such as ours might weigh context, free speech, the light tone, or other extenuating circumstances, the engineering state doesn't do that. It cannot self-correct mid-course. It is built to solve problems and the problem here is to protect heroes and martyrs, restore face, and deter future incidents. Partial retreat would leave the “problem” half-solved and look like incomplete execution. This is why the CCP proactively shapes historical memory and public sentiment. Mao’s status is non-negotiable. Tiananmen has to stay scrubbed from collective memory at all costs. Admitting the reaction was excessive or that the joke was harmless would concede ground on a foundational myth and imply the engineered social order is contestable. The funniest part is that the lengths they went to protect the narrative boundary meant they couldn’t even accurately report on what Mbappé said. “Chinese dictator” would itself be slandering a hero. So the Chinese people were told he had insulted China, with vague references to “inappropriate remarks.” Behind the Great Firewall, people are outraged and have purged Mbappé from everyday life while having no idea exactly why. It's basically like The Borg from Star Trek, operating as a hive-mind. Not to mention, "face" (面子) matters a lot in Chinese culture and the manic devotion to "saving face" makes about-turns almost impossible. This is why Mbappé’s, Javier Tebas’s, and Real Madrid’s apologies have zero impact. The engineering state acknowledged the apology (while still not being able to tell the Chinese people what it was for), but it will not meaningfully restore the contracts and licensing deals. The system is driven toward decisive, all-or-nothing execution on any problem it decides is important and to the Chinese, Mao is VERY important. Reversing course would be a public admission that the original solution was flawed. In a system that presents its policies as scientifically rational and historically correct, that’s moral and even civilizational suicide. That said, the system can be reversed. The one-child policy was eventually relaxed in 2015 and then ended. Zero-COVID was abandoned very abruptly in late 2022 after widespread protests (very rare in China), signaling that the costs of maintaining it was too high. What will likely happen here is what happened after the Chinese canceled the NBA and Houston Rockets over Daryl Morey's innocuous tweet supporting the Hong Kong pro-democracy movement in 2019. There will be no formal public reversal. Instead what you will see is a gradual and low-key restoration of the status quo. Daryl Morey left the Houston Rockets in 2020 and then by 2022, fully two and a half years later, Chinese state TV (CCTV) fully resumed regular NBA broadcasts. It was all very discreet. The lesson here is: never apologize. It's pointless. You only compromise your own integrity. The over-the-top reaction is a feature of the Chinese system, not a bug. It's not just because of different values, it's also because you're trying to do business with a culture that has implemented a specific institutional and cognitive style that treats society as yet another engineering problem to solve. There is no mechanism to self-correct so you just have to wait it out.
249
527
109
2,839
292,842
Ovidiu Ursachi 🇪🇺 retweeted
You misspelled “Russia”.
What’s actually going on in Ukraine is that there is a massive incentive difference between the elites and the people. The former have grown very rich off the hundreds of billions flowing to Ukraine, while the latter’s children are conscripted to die in trenches.
508
2,394
58
22,907
523,170
No, @elonmusk, the U.S. stopped funding Ukraine last year. In fact, U.S. companies make money on American weapons for which we Europeans pay. The more you meddle in our politics and the more you play politics with your products, the more you damage your brand.
Replying to @elonmusk
And some of that US taxpayer money comes back through multiple laundry machines to fund political campaigns in America, which is why you see some politicians caring more about Ukraine than America!
1,317
9,605
353
50,425
1,013,386
Ovidiu Ursachi 🇪🇺 retweeted
If confirmed, this is telling: Russian data centres get hit, and online activity around AfD suddenly plummets. 📉 So much talk about German patriotism… then the connection from Russia goes down. 🤔🇷🇺
1,593
11,889
1,609
62,701
4,360,429
Ovidiu Ursachi 🇪🇺 retweeted
„Zelensky has gone too far“ 🤡
62
354
13
11,414
331,740
Ovidiu Ursachi 🇪🇺 retweeted
Following a press investigation, the head of one of Romania’s most infamous and powerful intelligence services, and more or less President Nicușor Dan’s current right-hand man, was revealed to have been an officer of Ceaușescu’s secret police, the infamous Securitate. During the 1989 Revolution, he was personally in command of Securitate troops at a site where more than 48 civilians were executed, with hundreds more wounded, tortured, or arrested. In what is entirely consistent with what has been a disastrous first year in office, marked by a series of irrational decisions among which the refusal (still) to appoint a new PM, Nicușor Dan, a man that came to power promising reforms and to curb the influence of Romania’s secret services which have penetrated virtually every level of the country’s media, politics, and economy, has so far completely refused to take any action or even comment the matter. The fragile Romanian democracy, in no small part thanks to the actions of this president over which the Romanian society is currently split in thinking he is either insane or an agent of the secret services, is on the verge of completely collapse. However bad Hungary was for the EU, prepare for whatever damage an authocratic Romania that is twice the size and economy, holds a vital strategic position and is on the verge of becoming EU main gas exporter, will do to the EU.
45
251
21
1,029
50,786
Ovidiu Ursachi 🇪🇺 retweeted
Last week I posted a new paper with Patrick Norrick: “Terra Incognita: The Economics of a Shrinking World.” We chose the title deliberately. No society in history has experienced the fertility levels now seen in South Korea, China, Thailand, Colombia, Chile, and many other countries. Our knowledge of the causes (and of the economic consequences) remains far more limited than most public discussion acknowledges. Much of what we write is, at best, educated conjecture. The paper also struggles against a space limit. We wrote 20,000 words, far from the 250,000 or so we would need to address some issues in more detail (if I had the time and resources to hide away for a year, I could do that, but not now). That means some ideas are only sketched. Nonetheless, we emphasize several important points. First, fertility has fallen very fast everywhere: rich and poor countries, east and west, north and south, conservative and liberal societies, religious and non-religious societies (with the exception of the Jewish population of Israel; fertility has also collapsed among the Muslim population within the pre-1967 borders), you name it. Even in sub-Saharan Africa we see fast and unprecedented drops in fertility (alas, from a high initial level). Second, and this is really interesting, the fertility collapse has been concentrated in poor and lower-middle-income countries much more than in rich countries. By now, income per capita and fertility are positively correlated within OECD countries. We conjecture this will hold globally in a few decades. Third, and related to the second point, the fertility collapse has been concentrated among poor and lower-middle-income women within countries. In countries such as the U.S., the rich and highly educated now have more children than the poor and less educated. Fourth, we document why we do not understand the data from the U.N. World Population Prospects. See, for example, Tables A.1 and A.2. Fifth, we explain why some proposed mechanisms struggle when confronted with the data. A more subtle point is at work here. Many commentators do not seem to understand the difference between proximate causes and ultimate causes. Yes, births might have gone down because fewer women are in long-run relationships. That is the proximate cause. But you need to explain why fewer women are in long-run relationships (the ultimate cause), and saying that they spend more years in school, to take one example, does not get us very far. Why do they spend more years in school? Once you start down the whole chain of reasoning, things become much harder than they seem. The paper can be found here: sas.upenn.edu/~jesusfv/Demog… Comments always welcome!
294
1,545
320
8,993
2,798,561
Ovidiu Ursachi 🇪🇺 retweeted
Research paper finds: “Everyone gains with AI use in the short run. The long run depends on how you engage” The 8 Modes of AI Engagement: How You Work With AI Determines Whether It Makes You Sharper or Duller Most people treat AI as a smarter search engine or a tireless junior employee. They ask a question, receive a fluent answer, and move on. That pattern feels efficient. Research grounded in cognitive science, learning theory, and human-computer interaction suggests it is also quietly costly. The AI Engagement Modes framework at Brigham Young University) and detailed in the research foundations at aimodes.ai, classifies human-AI interaction into eight observable modes. These modes sit on a clear gradient of cognitive agency—from near-total passivity to full human control over problem framing. The framework draws on Bloom’s revised taxonomy, Chi’s ICAP framework (Interactive, Constructive, Active, Passive), Zimmerman’s model of self-regulated learning, Vygotsky’s Zone of Proximal Development, distributed cognition, epistemic vigilance, and Schön’s reflective practice. The central claim is straightforward: the way you engage AI shapes whether the technology amplifies your thinking or gradually erodes it. The eight modes are grouped into three tiers of increasing human agency: Passivity (Modes 1–2): The human largely cedes cognitive work to the AI. Partnership (Modes 3–4): Human and AI form a joint cognitive system with shared effort. Agency (Modes 5–8): The human retains or reasserts primary cognitive authority—verifying, challenging, expanding, or defining the problem itself. Higher tiers map to deeper learning and better long-term skill retention. Lower tiers map to short-term performance gains that often reverse when AI is removed. The 8 Ways of Engagement Here are the eight modes, ordered from lowest to highest cognitive agency: 1. Oracle You treat the AI as an authoritative source of answers. You ask broad or factual questions and largely accept the output. This corresponds to Bloom’s “remembering” level. It is efficient for quick facts but trains the habit of not checking or synthesizing. 2. Production (or Production Assistant) You use the AI primarily as a generator of finished or near-finished output—text, code, summaries, plans—with minimal editing, verification, or reframing. The cognitive work of creation is mostly offloaded. This is still passive in the sense that the human does little constructive or critical processing. 3. Tutor You use the AI as a scaffolded teacher. You ask for explanations, step-by-step guidance, or clarification within your Zone of Proximal Development. You remain engaged enough to learn rather than merely receive. This is the first genuine partnership mode. 4. Collaborative Problem-Solver You and the AI work together as a distributed cognitive system. You break problems into steps, iterate, supply context, and build on the AI’s contributions while directing the overall direction. Cognition is shared rather than handed over. 5. Verification Agent You actively evaluate the reliability of the AI’s claims before accepting them. You check assumptions, cross-reference, test edge cases, or demand evidence. This reactivates epistemic vigilance—the evolved capacity to scrutinize communicated information that fluent AI output can otherwise bypass. 6. Critical Challenger You treat the AI as a sparring partner. You argue against its reasoning, force it to defend positions, surface weaknesses, or explore counter-arguments. This leverages the argumentative nature of human reasoning and produces stronger evaluation skills. 7. Creative Expander You use the AI to push beyond the given frame—generating novel variations, analogies, alternative framings, or divergent possibilities. The human remains the director of creative direction while the AI multiplies options. 1 of 2
47
138
37
809
50,868
Ovidiu Ursachi 🇪🇺 retweeted
New: A map of the most important skills in AI Engineering.
592
4,146
485
24,674
7,051,356
Ovidiu Ursachi 🇪🇺 retweeted
There's a major gap in this otherwise compelling vision. Lots of interesting signals recently: token budget caps, intelligence sovereignty, increasing competition (Meta/Grok/GLM) that drives the commoditization of intelligence. All seem to point toward a more benevolent future equilibrium where every organization owns its human-AI learning loop through which knowledge and value accrue locally. @satyanadella, @thinkymachines, @dwarkesh_sp all laid out good arguments for it. No doubt that is a more desirable future than the oligopolistic path we are on, but no one seems to be asking about the elephant in the room: Do we actually know how to solve the continual learning problem required to sustain such a local co-learning loop? Right now, there are roughly three buckets of learning methods being entertained in practice: 1) Non-parametric learning, mostly in some form of memory (skills, RAG, knowledge bases, KV cache, etc) 2) Domain-level post-training (e.g., continual pre-training on proprietary organizational data) 3) Task-level post-training (e.g., RL for specific workflows) (More research-y ones like new model architectures are omitted here because there's still a long way to go in both research and validation) Is any of these methods sufficiently general and deep to sustain the desired co-learning loop across all organizations and job functions? The answer is likely negative: > Non-parametric memory is often shallow and has limited control over agent behavior (talk to OpenClaw/Hermes users who struggle to get their agents to learn and follow the rules) > Domain-level post-training remains expensive and has yet to demonstrate broad success outside a few exceptional domains (@Cursor's Composer may be an exception but coding is an exceptional domain in itself) > Task-level RL is engineering-heavy, sample-inefficient, and difficult to apply when success cannot be objectively verified The human-AI learning loop isn’t inevitable. It still needs to be invented. Solving continual learning may be one of the most important problems for building an AI ecosystem where expertise compounds locally instead of concentrating globally.
13
19
3
164
30,444
Ovidiu Ursachi 🇪🇺 retweeted
Our statement on the UK government’s demand that all content on all devices sold or used in the country be scanned, on the presumption of nudity, using a dystopian combination of age verification and content scanning. This proposal will not safeguard children. It endangers us all. signal.org/blog/pdfs/2026-06…
716
8,214
501
40,272
2,957,238
Chapeau @AlexZverev für den Titel beim Roland-Garros! Es ist mehr als verdient.
29
Ovidiu Ursachi 🇪🇺 retweeted
Doamna Buzo lovește din nou! Femeia asta are cojones cât 1000 de putiniști/pesediști. „100 de hectare exploatate ilegal în Olt. Ce zic cei responsabili să oprească asta? Că nu au văzut nimic. Aceste firme, atât de bine cotate, și-au garantat profitul sărăcind pământul nostru, într-un mod ilegal, care scoate la iveală un nivel uriaș al corupției și al complicității. Cele două companii au fost deja sancționate, iar demersurile noastre continuă. Cei care trebuiau să facă verificări, să dispună încetarea lucrărilor și nu au făcut-o vor fi, la rândul lor, trași la răspundere. Totul conform legii. Pentru că legea trebuie să fie mai puternică decât orice rețea de interese”. (Diana Buzoianu / FB)
19
72
2
601
6,243
Ovidiu Ursachi 🇪🇺 retweeted
I don't remember where I found this, but its spot on.
715
6,325
765
31,147
45,167,619
Ovidiu Ursachi 🇪🇺 retweeted
Nicușor Dan, the president Romanians elected last year precisely to prevent anti-EU extremist forces from taking over, has turned out to be one of the main players behind the ousting of the pro-EU reformist PM Bolojan, as best we can tell simply out of personal dislike and to eliminate a very popular potential future rival. Predictably, the only likely outcome at this point seems to be a governing coalition that will include, in some shape or form, both the kleptocratic socialists and the pro-Russian, anti-EU, Iron Guard wannabe extremists from AUR, led by George Simion, a man barred from entering the territory of Ukraine and Moldova due to his connections to Russian intelligence. I do not think I can do justice to the anger and sense of betrayal felt throughout Romanian society right now, all directed towards Nicusor Dan, a man that out of pettiness and sheer stupidity has likely just paved the way to power to a pretty much openly fascist party.
49
141
17
688
27,713
Replying to @KLM
@KLM how is it possible that I book a ticket on your own app, I want now to add a luggage but this is not doable because I am flying first leg with @transavia which belongs to same group and I must pay tomorrow at the airport huge fees for the luggage?
9
2
120
How is it possible that I cannot checkin on your own app and I must checkin on @transavia . If I do not do this, I have to pay checkin fee at the airport. If I do the checkin there now, I have to pay luggae fee for one leg at the airport
2
1
1
1
90
Ovidiu Ursachi 🇪🇺 retweeted
I packaged up the "autoresearch" project into a new self-contained minimal repo if people would like to play over the weekend. It's basically nanochat LLM training core stripped down to a single-GPU, one file version of ~630 lines of code, then: - the human iterates on the prompt (.md) - the AI agent iterates on the training code (.py) The goal is to engineer your agents to make the fastest research progress indefinitely and without any of your own involvement. In the image, every dot is a complete LLM training run that lasts exactly 5 minutes. The agent works in an autonomous loop on a git feature branch and accumulates git commits to the training script as it finds better settings (of lower validation loss by the end) of the neural network architecture, the optimizer, all the hyperparameters, etc. You can imagine comparing the research progress of different prompts, different agents, etc. github.com/karpathy/autorese… Part code, part sci-fi, and a pinch of psychosis :)
1,042
3,571
1,136
28,109
11,182,766
Ovidiu Ursachi 🇪🇺 retweeted
I Built an Interactive Voice-Narrated Masterclass From My Book in One Weekend... No studio. No production team. No six-month timeline. One AI workflow that turned a 300-page book into a live, interactive learning product. Here’s the pipeline I used: 1️⃣ Mega-prompt blueprint: I treated the prompt like architecture, not a “summarize my book” request. 2️⃣ Book → 8-episode learning arc: Claude mapped the chapters into a coherent transformation journey, ending in a concrete 90-day plan built around WAVE. 3️⃣ Interactive HTML episodes: 8 web-native presentations with animations, quizzes, and progress tracking. 4️⃣ 13,000+ words of narration scripts: written to match my speaking style and synced to slide reveals. 5️⃣ Audio + slide sync: ElevenLabs voice cloning connected to Claude Code via MCP, auto-generating narration and perfectly timed slide progression. 6️⃣ Deploy: gated enrollment + Ghost CMS, one command to go live. This is the real shift: AI doesn’t replace expertise, it removes the production bottleneck between deep knowledge and scalable delivery. The masterclass is available for free, and a detailed explanation is on my website. No need to comment/DM me for access... Masterclass: thedigitalspeaker.com/book-n… Detailed blog: thedigitalspeaker.com/book-i…
2
3
274