Info
Building web and back-end high performance applications 💻. Broad experience in…
Aktivitäten
6149 Follower:innen
-
Slava Rosin hat dies repostetSlava Rosin hat dies repostetAnother great moment for Estonia’s AI Leap (TI-Hüpe) - this time featured on CNN: https://lnkd.in/dF4PjSmt Minister Kristina Kallas, Ministry of Education and Research of Estonia / Haridus- ja Teadusministeerium spoke about #Estonia’s approach to bringing #AI into schools: not banning it, but learning how to use it in a way that strengthens students’ thinking and helps education evolve for the AI era. A few points from the CNN conversation highlighted by Minister Kristina Kallas: • The focus is not simply on access to AI, but on understanding what kind of learning happens when AI becomes part of the learning process. • Teachers have shown strong interest in participating, and the initiative has received a very positive response from schools and families. • Estonia’s 25+ years of experience with digital education gives us a strong foundation for experimenting with the next generation of learning tools. • The ambition goes beyond efficiency: the goal is to explore whether AI can support an “evolutionary jump” in students’ intellectual capabilities and encourage much deeper thinking. For colleagues who would like to go deeper, we’ve also published a report covering the first-year results and learnings from AI Leap (in English) 👉: https://lnkd.in/dgTEhBKU It’s exciting to see Estonia’s experiment getting international attention - but even more exciting to see the evidence and learning starting to emerge from the work itself. #AILeap #ArtificialIntelligence Askend Targa Tuleviku Fond / Smart Future Foundation #Education Praktikal #EdTech #Estonia #FutureOfEducation
-
Slava Rosin hat dies repostetSlava Rosin hat dies repostetDear European Central Bank, we went ahead and finished your redesign for you 💶 Everyone's slapping their faces on euro notes this week. So we did it properly, with the people who actually built modern Europe 🇪🇺 Three small countries. One very good reason to reprint 📈 No thanks needed, ECB. Let's build 💜 Markus Villig, Milda Mitkutė, Justas Janauskas, Tomas Okmanas, Eimantas S., Davis Siksnans, Kristo Käärmann, Taavet Hinrikus, Ahti Heinla, Priit Kasesalu, Jaan Tallinn #FIRSTPICK #vc #startups #Baltics #founders
-
Slava Rosin hat dies repostetSlava Rosin hat dies repostetvibe coders after they hit their daily usage limit
-
Slava Rosin hat dies repostetSlava Rosin hat dies repostetA founder trusted their developer for months. Then we opened the repo. The code was barely checked in, and critical components were unimplemented and mocked out. Nothing another engineer could pick up. That gap crippled a launch and forced a scramble to reset every date. And that's the obvious problem. The subtle one costs more: a builder shipping against today's APIs, and no one above them asking whether those shortcuts survive next year (or next month), especially now that hyperscalers wipe out entire layers of APIs overnight. In the age of AI, you don't just need someone to wield the tools and ship an MVP. You need someone thinking ahead of the immediate implementation: scale, data, compliance, and where every tool adds friction or slows your path to customers. That's the case for a fractional CTO. I made it in full in this week's Escape Velocity. Where do you need a nudge on something your developers aren't thinking about? Full piece below.Your Developer Is Not Your CTO: Why Startups Need High-Quality AdvisorsYour Developer Is Not Your CTO: Why Startups Need High-Quality Advisors
-
Slava Rosin hat dies repostetSlava Rosin hat dies repostetFable 5 is back. The frontier model the US government pulled from the market last month returned on June 30, and Anthropic has finally explained what actually happened. The resolution is more revealing than the suspension was. The key admission: Anthropic tested the reported "jailbreak" against weaker models, including Opus 4.8, GPT-5.5, and Kimi K2.7, and every one of them could find the same vulnerabilities. It unlocked no unique capability, it just tripped a deliberately cautious safety margin. The fix is a new classifier that blocks the technique in over 99% of cases, at the cost of more false positives on ordinary coding and debugging. So the model is back, a little more restricted, and the security gap that triggered a worldwide suspension turned out to be narrow. The part that outlasts this episode: the US government is now in the release loop for frontier models, and Anthropic is drafting a shared jailbreak-severity standard with Amazon, Microsoft, and Google. That's the right instinct, but it still doesn't answer the harder question of who decides when a deployed model is too risky to run, and on what evidence. For anyone building on these models, capability is no longer the only variable. Availability and regulatory posture now belong in your planning too.
-
Slava Rosin hat dies repostetSlava Rosin hat dies repostet𝐖𝐡𝐚𝐭 𝐢𝐬 𝐛𝐞𝐭𝐭𝐞𝐫 𝐭𝐡𝐚𝐧 𝐡𝐚𝐯𝐢𝐧𝐠 𝟏𝟎𝟎 𝐟𝐨𝐮𝐧𝐝𝐞𝐫𝐬, 𝐢𝐧𝐯𝐞𝐬𝐭𝐨𝐫𝐬 𝐚𝐧𝐝 𝐕𝐂𝐬 𝐭𝐨𝐠𝐞𝐭𝐡𝐞𝐫 𝐟𝐨𝐫 𝟏 𝐰𝐞𝐞𝐤? Having all of that in Italy, on the beach! 🏖️🇮🇹 And by now, you all know: I love the beach. And I especially love talking business on the beach. ☀️ So it is my absolute pleasure to bring my unconventional perspectives on entrepreneurship and my high-vibrational energy all the way to Puglia, Italy. I'll be speaking about my next favourite topic: Resilience for Founders. 𝐁𝐞𝐜𝐚𝐮𝐬𝐞 𝐲𝐨𝐮 𝐜𝐚𝐧 𝐨𝐧𝐥𝐲 𝐠𝐫𝐨𝐰 𝐲𝐨𝐮𝐫 𝐛𝐮𝐬𝐢𝐧𝐞𝐬𝐬 𝐚𝐬 𝐦𝐮𝐜𝐡 𝐚𝐬 𝐲𝐨𝐮 𝐠𝐫𝐨𝐰 𝐲𝐨𝐮𝐫𝐬𝐞𝐥𝐟. 💛 Thanks to Lorenzo from Feedel Ventures and Adam from European Startup Week for the invite! And beyond excited to be sharing the stage with some of Europe's finest unicorns 🦄 ◆ Dmitry Gurski - Co-Founder & CEO, Flo Health Inc. ◆ Razvan Ranca - Co-Founder & CTO, Tractable ◆ Philipp Nieland- Co-Founder, PPRO 📅 6 July - 12 July 2026 📍 Puglia, Italia Thank you for enabling this gathering to the Minister for Economic Development, Eugenio Di Sciascio and the incredible partners: Emilia from Deep Ocean Capital SGR S.P.A., Annamaria from Angels4Women, Mike from Profondo Capital & Co-Owner, Cesena Football Club. And so excited to be joining such an incredible lineup: Salvatore, Paolo, Max, Fabrizio, Rob, Bibop, Sara, Dominic, Ken, Patrick, Neal, Leonne, Claudio, Jason, Fabiana, Alfredo, Massimiliano, Kevin. Counting the days. See you soon - on the beach! 🏖️ #EuropeanStartupWeek #EuropeanStartupWeek2026 #Entrepreneurship #Startuplife #Founderslife #FounderResilience
-
Slava Rosin hat dies repostetSlava Rosin hat dies repostetOver the last 18 months I have come to one conclusion. The biggest mistake people make about AI is not fear, it is pretending. Pretending that nothing has changed, pretending that this is just another technology cycle, pretending that the next decade will look roughly like the last one. It won't. I do not know exactly what the world will look like in 10 years. I do not know which companies will win. I do not know which professions will disappear. I do not even know which tools we will be using 2 years from now. But a few things seem obvious. AI gets better every week. The cost of intelligence is falling. The amount one person can accomplish is rising, and the pace of change is accelerating. So my conclusion is simple: Don't pretend. Don't pretend the world has not changed. Don't pretend the skill does not exist. Don't pretend the next 10 years will look like the previous 10. The changes are already here. The question is no longer whether they will happen. The question is, what will each of us choose to do next. Built for Better Decisions.
-
Slava Rosin hat dies repostetThank you for the kind words! 🤝Slava Rosin hat dies repostetFIBO 2026 ✓ Köln, gute Gespräche und ein Abend mit dem Team von AxlBrains. 🤝 Danke für den lieben Austausch, die Gesellschafft und die Einblicke! Checkt doch mal die @axl.coach website, tolles Team & tolles Produkt! #FIBO2026 #companycore #axl #FitnessBranche #KI
-
Slava Rosin hat dies repostetSlava Rosin hat dies repostetToday I trained the Prime Minister and the Secretary of State of Estonia on how to use AI agents. Do you know what the PM built? We started with the latest agentic developments. I showed how Pactum AI agents operate end to end, autonomously and alongside humans. I think it was useful for them to see real agents in action, not just the buzz of wrappers. Then I walked through how Claude Code is used across Pactum's different roles. We've trained it through custom skills and have given access it into all our key systems. The specific examples stood out from HR, Ops, to Legal all have each set up their own agents on Claude to help run their operations. Then the PM took over and started vibe coding on Claude Code and Lovable himself. He built a "PM Cockpit" that pulled together all key government priorities with existing progress reports and escalation points. In parallel, he asked Claude to build it into a slide deck for the next cabinet meeting. And in parallel, he asked Claude to prep for a press conference by generating the most difficult media questions and draft answers for the same topics. All of this in one sitting, multitasking. It helps when you continuously have leaders who just get it. Who want to get their hands dirty. Who keep learning and building. Can you imagine your Prime Minister setting up agents to help run your country? Running a company is difficult, but running a country is next level.
Slava Rosin hat einen Beitrag kommentiert
1 Monat
Baruch is legend
-
Slava Rosin gefällt dasSlava Rosin gefällt dasFor about a year I used AI to get through more work. That was the wrong instruction to give myself. More got through. The desk emptied earlier, the replies went out the same day, and by any measure a consultant would put in a deck the arrangement was working. What the deck would not show is that the number of things I was personally deciding had gone up in step with the speed, and a person only has so many good decisions in a day. I had bought speed with the one thing I should have been protecting. The principle took an embarrassing amount of evidence to reach. A tool that makes each item on the desk cheaper to handle will increase the number of items you handle, unless you point it at the opposite job. So the first instruction each morning is subtraction. Not summarise this. Tell me whether this needs a CEO. The tool reads the queue before I do, argues for what I can leave shut, and gives its reason. I read the reasons. I once tracked every input that crossed my screen for a week. More than 200 of them. About 20 changed what I actually did. Roughly one in ten, and I did not believe it at first. The tool is good at finding that ratio, because it has no social reason to pretend an email matters just because a senior person sent it. It is not reliably right. It underweights the quiet item with no deadline, which is often the start of something that matters in a year, so I still open more than it recommends. I want to be clear about one thing, because the language around this gets loose. The machine reads, sorts, drafts and disagrees with me. It does not run the company and it does not decide. The choice stays with a person who can be held to it. If I put a commitment in writing for later, it is this: I would like our own use of these tools judged by whether the number of decisions concentrated at the top of the company goes down, and I would rather be measured on that than on any productivity figure. If anyone is already tracking that in their own company, I would rather borrow your count than invent one.
-
Slava Rosin gefällt dasSlava Rosin gefällt dasOur FORTALESA #chips arrived 🔬 FORTALESA is a run-time fault-tolerance reconfigurable #systolic_array for #DNN inference: three execution modes to balance throughput and #reliability layer by layer, based on each DNN layer’s criticality. The idea, architecture and reliability assessment method were published by 𝐍𝐚𝐭𝐚𝐥𝐢𝐚 𝐂𝐡𝐞𝐫𝐞𝐳𝐨𝐯𝐚, a PhD student in our #TECH team, and have now been turned into silicon. RTL, #OpenSourceEDA flow, 10mm² in IHP 130 nm BiCMOS #OpenPDK, September 2025 OpenMPW shuttle… A first tapeout is hard. A first tapeout on a fully open stack is harder. And it became possible only with advice from Frank Kagan Gurkaynak (ETH Zürich) and the support of the #IHP team, who gave us the PDK, the shuttle, and the patience. Both are partners in the #HorizonEurope project TAICHIP Twinning Project, supporting TalTech in boosting its chip-design capacity. The open chip package isn't for photos. It is for the irradiation and accelerated ageing tests that come next. The design is a step toward our TAICHIP #EdgeAI SoC platform. The next chip, a #RISCV + FORTALESA-based AI-SoC, was already taped out a couple of weeks ago. FORTALESA chip: https://lnkd.in/dgWQ-zCz TAICHIP AI-SoC: https://lnkd.in/dDmVxweP Congratulations, Natalia! #OpenSourceSilicon, #OpenMPW, #yosys, #openroad, #klayout TalTech – Tallinn University of Technology, IHP – Leibniz Institute for High Performance Microelectronics, European Research Executive Agency (REA)
-
Slava Rosin gefällt dasSlava Rosin gefällt dasWe’re excited to share that Versonix is moving to Los Gatos, California, effective September 1, 2026. 📍 720 University Avenue, Suite 130 Los Gatos, CA 95032 We’re looking forward to settling into our new space and continuing to build what’s next with our team, customers, and partners.
-
Slava Rosin gefällt dasSlava Rosin gefällt dasDuring university I worked as a waitress at Hard Rock Cafe in Copenhagen. People assume it's a fun but throwaway job. It wasn't. To this day it's one of the best examples I've seen of management systems, staff training, team culture, and performance culture done right. They were serious about it. And they made us SELL. Actively. Every table, every round. (Sell = bacon on a veggie burger type of selling 😃.) The principle I taught myself: if you don't ask, you won't sell. You just have to give every customer the chance to say yes. I've carried that into my later jobs since. It looks different: a client on the other side of Europe, a service that's not exactly your core offering, an industry you don't know well. But if there's a genuine opportunity to help, and you actually believe you could, ask. There's a LOT of talk about prioritisation. And yes, you have to do it. But there's a real art to staying in the game with minimum effort in places that aren't your strategic core. You never know which conversation becomes the one that matters.
-
Slava Rosin gefällt dasSlava Rosin gefällt dasEstonia Isn't Just Talking About the AI Education Problem. They're Building for It. One thing from my previous post kept bothering me: "If AI is taking over more of the lower-order thinking, what does education need to do next?" I came across something that made me think about that question differently. Estonia is already experimenting with an answer. Estonia has launched AI Leap, a national initiative bringing AI into secondary and vocational education. The scale is already significant: • 20,000 students and 3,000 teachers in the first phase • Plans to expand to 58,000+ students • Around €4 million in funding • Collaboration involving OpenAI and Estonian education institutions But what impressed me most wasn't the scale It was the philosophy behind it: ->The idea is to use AI as a “smart coach, not a smart servant.” That distinction matters. Instead of asking AI for an answer and copying it, students are encouraged to use it to work through the reasoning and reach the answer themselves. And that's exactly where the conversation about AI and education gets interesting. If AI can increasingly handle remembering, retrieving, and applying information, then perhaps education needs to put more emphasis on what comes next: 🔵Analyzing. 🔵Evaluating. 🔵Creating. And this isn't only relevant to students. Think about SQL, Excel, Python, or even AI itself. Knowing how to use the tool is becoming easier. The harder and more valuable part is knowing: → What question should I ask? → Does this result actually make sense? → What assumptions am I making? → What information is missing? → What decision should I make from this? Thats where human judgment starts becoming more important, not less. What I find particularly interesting about Estonia's approach is that they are not treating AI simply as a technology to introduce into classrooms. They are experimenting with how learning itself should change because AI exists. There are also efforts around data protection, research, and studying how AI affects students thinking and attention over time. That makes this more than just: “Lets put AI in schools.” Its closer to: “Lets figure out what education should look like when AI is already part of the learning environment.” For countries like Pakistan, I think this is an important conversation to start having now. Not: “Should students use AI?” But: “What should students learn when AI is already there?” What do you think? Should more countries follow Estonias approach, or is education moving toward AI faster than it can adapt? #AI #Education #ArtificialIntelligence #DataAnalytics #CriticalThinking #Pakistan #LearningInPublic
-
Slava Rosin gefällt dasSlava Rosin gefällt dasAnother great moment for Estonia’s AI Leap (TI-Hüpe) - this time featured on CNN: https://lnkd.in/dF4PjSmt Minister Kristina Kallas, Ministry of Education and Research of Estonia / Haridus- ja Teadusministeerium spoke about #Estonia’s approach to bringing #AI into schools: not banning it, but learning how to use it in a way that strengthens students’ thinking and helps education evolve for the AI era. A few points from the CNN conversation highlighted by Minister Kristina Kallas: • The focus is not simply on access to AI, but on understanding what kind of learning happens when AI becomes part of the learning process. • Teachers have shown strong interest in participating, and the initiative has received a very positive response from schools and families. • Estonia’s 25+ years of experience with digital education gives us a strong foundation for experimenting with the next generation of learning tools. • The ambition goes beyond efficiency: the goal is to explore whether AI can support an “evolutionary jump” in students’ intellectual capabilities and encourage much deeper thinking. For colleagues who would like to go deeper, we’ve also published a report covering the first-year results and learnings from AI Leap (in English) 👉: https://lnkd.in/dgTEhBKU It’s exciting to see Estonia’s experiment getting international attention - but even more exciting to see the evidence and learning starting to emerge from the work itself. #AILeap #ArtificialIntelligence Askend Targa Tuleviku Fond / Smart Future Foundation #Education Praktikal #EdTech #Estonia #FutureOfEducation
Berufserfahrung und Ausbildung
Bescheinigungen und Zertifikate
Veröffentlichungen
-
Trainer 1149: a Boundary Scan Simulation Bundle for Labs
18th International Conference Mixed Design of Integrated Circuits and Systems
-
Understanding Boundary Scan Test with Trainer 1149
The 22nd EAEEIE Annual Conference
In this paper, we describe a training environment based on multi-functional software system called “Trainer 1149”. It provides simulation and demonstration functionality for learning, research, and development related to IEEE 1149.1 Boundary Scan (BS) standard. The software supports both the analytic and synthetic learning process. Trainer 1149 is the main component of a recent goJTAG initiative that aims at creating of an open-source platform for learning and working with BS. Trainer 1149…
In this paper, we describe a training environment based on multi-functional software system called “Trainer 1149”. It provides simulation and demonstration functionality for learning, research, and development related to IEEE 1149.1 Boundary Scan (BS) standard. The software supports both the analytic and synthetic learning process. Trainer 1149 is the main component of a recent goJTAG initiative that aims at creating of an open-source platform for learning and working with BS. Trainer 1149 provides a cozy graphical design and simulation environment of BS-enabled chips and non-BS clusters. It is also possible to simulate the behaviour of various interconnect faults and inspect them using interactive tools. Using a convenient low-cost USB-JTAG cable, one can test real defects in real hardware. The system is implemented using multi-platform Java environment and distributed as an open-source freeware. Such combination of features is unique for public domain BS software.
Andere Autor:innenVeröffentlichung anzeigen
Kurse
-
Algorithms: Design and Analysis, Part I by Tim Roughgarden
-
-
Functional Programming Principles in Scala by Martin Odersky
-
-
Java Performance Tuning Course by Kirk Pepperdine
-
Auszeichnungen/Preise
-
Munich Half-Marathon
-
Sprachen
-
Russian
Muttersprache oder zweisprachig
-
English
Verhandlungssicher
-
Estonian
Verhandlungssicher
-
German
Fließend
Erhaltene Empfehlungen
6 Personen haben Slava Rosin empfohlen
Jetzt anmelden und ansehen