Advisor & Entrepreneur https://jannis.ritterspach.mehttps://bsky.app/profile/jannis.ritterspach.mehttps://bsky.app/profile/jannis.ritterspach.me/post/3msdcivsz5226MIT CISR on why AI projects fail: organizations apply yesterday's best practices to an inherently different technology. My take: leaders treat AI as a vertical. A project, a tool, something the innovation team owns. It's horizontal. Rethinking how you operate is the AI strategy.05 Aug 2026 10:20 +0000at://did:plc:kaiath4dugtus25pxstdzv2y/app.bsky.feed.post/3msdcivsz5226https://bsky.app/profile/jannis.ritterspach.me/post/3msbb6aj2qs2pStanford HAI on the market for "sovereign AI": these offerings reconfigure dependencies rather than removing them. Aleph Alpha, Germany's sovereign LLM champion, now belongs to Canadian Cohere. Better question: what would it cost me to leave, and how long would it take?04 Aug 2026 14:50 +0000at://did:plc:kaiath4dugtus25pxstdzv2y/app.bsky.feed.post/3msbb6aj2qs2phttps://bsky.app/profile/jannis.ritterspach.me/post/3ms64achwt22rBCG's first Intelligent Cities Index ranks 61 cities on how they use AI. Eight European cities in the top 20, Berlin sixth. The part worth noting: no city performs well on soft enablers (talent, funding) alone. Infrastructure turns out to be the easier half.03 Aug 2026 08:44 +0000at://did:plc:kaiath4dugtus25pxstdzv2y/app.bsky.feed.post/3ms64achwt22rhttps://bsky.app/profile/jannis.ritterspach.me/post/3mri7jczzo22gAfter 24 hours with the Opus 5, I can say with confidence that this model is a beast. The coding ability is impressive. It's not just incrementally better; it's noticeably better at holding context across a long codebase and sessions. It's my favorite model by far right now.25 Jul 2026 15:44 +0000at://did:plc:kaiath4dugtus25pxstdzv2y/app.bsky.feed.post/3mri7jczzo22ghttps://bsky.app/profile/jannis.ritterspach.me/post/3mr6ivjdlws23Six months ago my GitHub graph showed 3,462 contributions. Today it's 10,021. I haven't written a line of code in nine months. It was an experiment: how far can one person with AI go in depth, not just demos? Answer so far: several production systems. Next question: how to scale it horizontally.21 Jul 2026 19:06 +0000at://did:plc:kaiath4dugtus25pxstdzv2y/app.bsky.feed.post/3mr6ivjdlws23https://bsky.app/profile/jannis.ritterspach.me/post/3mqs6myzpu22cYour brain is lying to you about AI. It thinks in straight lines. AI moves on an exponential. METR's task-length horizon doubled every 7 months for six years; recently every four. GPT-2: two seconds. Claude 3.7: 50 minutes. Opus 4.6: twelve hours. Our intuition is linear. The curve is not.16 Jul 2026 21:30 +0000at://did:plc:kaiath4dugtus25pxstdzv2y/app.bsky.feed.post/3mqs6myzpu22chttps://bsky.app/profile/jannis.ritterspach.me/post/3mqphh7rhh22qThe team behind AI 2027 just published a sequel: AI 2040, or "Plan A." The first was bleak: a race to superintelligence ending in extinction or extreme concentration of power. Plan A is the counterproposal; a US-China deal to slow the race on purpose.15 Jul 2026 19:30 +0000at://did:plc:kaiath4dugtus25pxstdzv2y/app.bsky.feed.post/3mqphh7rhh22qhttps://bsky.app/profile/jannis.ritterspach.me/post/3mqoj4l5mas2pAI integration > tool usage. New HBR research: 199 founders on what AI did to their headcount. Without it, they'd need 55% more staff on average. But the median was 17%, and the most common answer was zero. The difference wasn't AI's capability. It was whether they'd integrated it into workflows.15 Jul 2026 10:27 +0000at://did:plc:kaiath4dugtus25pxstdzv2y/app.bsky.feed.post/3mqoj4l5mas2phttps://bsky.app/profile/jannis.ritterspach.me/post/3mq7spenvo223Nearly 50% of all AI agent activity happens in one domain: software engineering. Everything else sits in the low single digits. The easy reading: those fields aren't affected yet. I read it differently. That's the before picture; a preview of what's coming for knowledge work as a whole.09 Jul 2026 14:09 +0000at://did:plc:kaiath4dugtus25pxstdzv2y/app.bsky.feed.post/3mq7spenvo223https://bsky.app/profile/jannis.ritterspach.me/post/3mpyvqk66x22zThe EMBA in numbers. 16 modules. 32 written assignments. 18 train trips from Berlin to Lausanne and back; ~30,000 km on the rails, almost three-quarters around the planet. 342 hours in a train seat. And 140 pages of thesis at the end. Done. And already thinking about what's next.06 Jul 2026 20:14 +0000at://did:plc:kaiath4dugtus25pxstdzv2y/app.bsky.feed.post/3mpyvqk66x22zhttps://bsky.app/profile/jannis.ritterspach.me/post/3moguuoxln22oNew WordPress VIP report, one line holds it together: brands chase AI visibility, consumers chase the source. 74% of leaders prioritize AI discoverability. Yet 42% of consumers trust unattributed AI answers less than medical bills and airline fees. wpvip.com/future-of-the-web-202616 Jun 2026 22:46 +0000at://did:plc:kaiath4dugtus25pxstdzv2y/app.bsky.feed.post/3moguuoxln22ohttps://bsky.app/profile/jannis.ritterspach.me/post/3mo4q2kurcs2eAnthropic published its progress on recursive self-improvement. AI building AI. Claude now writes 80%+ of the code in its own codebase. As AI takes over execution, the human role narrows to judgment. The "doing" now costs almost nothing. What remains is taste. https://www.anthropic.com/institute/recursive-self-improvement12 Jun 2026 21:53 +0000at://did:plc:kaiath4dugtus25pxstdzv2y/app.bsky.feed.post/3mo4q2kurcs2ehttps://bsky.app/profile/jannis.ritterspach.me/post/3mnxescxl4s2hOne Europe. One Standard. The EU-INC proposal can work, but only if the free choice of registration seat stays in. Without it, we get 27 national regimes sharing a logo. July 16 is the deadline. This is Europe's startup moment; don't let it get watered down. eu-inc.org10 Jun 2026 18:48 +0000at://did:plc:kaiath4dugtus25pxstdzv2y/app.bsky.feed.post/3mnxescxl4s2hhttps://bsky.app/profile/jannis.ritterspach.me/post/3mnxbxo2hus2pI have advised Adiuva Capital since 2024, under the radar. Time to say it out loud. We work on a buy-and-build initiative in DACH digital services: platform strategy, sourcing, and target evaluation. This is the work I care about most: turning fragmented expertise into something that works as one.10 Jun 2026 17:57 +0000at://did:plc:kaiath4dugtus25pxstdzv2y/app.bsky.feed.post/3mnxbxo2hus2phttps://bsky.app/profile/jannis.ritterspach.me/post/3mjmzfhxaxc2mStanford 2026 AI Index: SWE-bench coding went from 60% to near 100% in one year. US-China gap effectively closed. Gen AI hit 53% adoption in 3 years – faster than PC or internet. US ranks 24th at 28%. https://hai.stanford.edu/ai-index/2026-ai-index-report16 Apr 2026 18:17 +0000at://did:plc:kaiath4dugtus25pxstdzv2y/app.bsky.feed.post/3mjmzfhxaxc2mhttps://bsky.app/profile/jannis.ritterspach.me/post/3mjl4bz4oy22v65% of agency jobs may not survive AI. Marketing jobs down 7% YoY, junior roles first. The real challenge isn't adopting tools. It's building organizations where AI works as a workforce, not just a tool. https://www.adweek.com/brand-marketing/65-of-marketing-jobs-may-not-survive-ai/16 Apr 2026 00:03 +0000at://did:plc:kaiath4dugtus25pxstdzv2y/app.bsky.feed.post/3mjl4bz4oy22vhttps://bsky.app/profile/jannis.ritterspach.me/post/3mjia3mozrs2aAnthropic research: AI exposure varies widely. Programmers at 75%, customer service at 70%, data entry at 67%. The gap between what AI could automate and what's actually happening is huge. It's about adoption, not capability. https://www.anthropic.com/research/labor-market-impacts14 Apr 2026 20:33 +0000at://did:plc:kaiath4dugtus25pxstdzv2y/app.bsky.feed.post/3mjia3mozrs2ahttps://bsky.app/profile/jannis.ritterspach.me/post/3mfoftidzws2c@hbr.org research: Marketing faces two disruptions. The first (how we search) gets all the attention. The second matters more: bots as customers. When AI handles research and transactions, the customer journey happens inside an algorithm. No website visit, no brand impression, no conscious choice. https://hbr.org/2026/02/ai-is-upending-marketing-on-two-fronts25 Feb 2026 09:51 +0000at://did:plc:kaiath4dugtus25pxstdzv2y/app.bsky.feed.post/3mfoftidzws2c