Donald Gossen

The portuguese XR Market essentially hires across the world and remote work is no longer a differentiator but a commodity. Know more about it in the Portugal XR Report 2025, available for download here: https://lnkd.in/dtUbQf-9

15

Oraion has raised $3.5M in pre-seed funding for its agentic AI platform. Oraion's AI-powered enterprise platform provides organizations a direct line from raw data to clear, actionable decision-making. It integrates seamlessly with productivity applications like Slack and Microsoft Teams, allowing executives to interact directly with Oraion through their existing tools and receive deep insights without workflow disruption. The platform supports over 300 data sources. The investment was led by Studio VC, with participation from Wayflyer, Indeed founder and YouTube's principal data scientist, Angus Miln (Cooley LLP), Maria O’Brien (SOSV), and Pierre-Antoine Porte (OpenAI). The $3.5M pre-seed investment will enable Oraion to triple its workforce to 45, accelerate product development, and expand into the US and Latin America, with a go-to-market target of 2026. Congratulations to co-founder and CEO Alexander Walsh and the Oraion team. Silicon Republic: https://lnkd.in/dZPzKtEE #AI #AgenticAI #EnterpriseAI #DataAnalytics #Productivity #PreSeed #Funding #IrishTech #USTech

13

Why do 98% of AI agent pilots never make it to production? Most teams think scaling is just “connect some APIs and call it a day.” Reality check: true scale demands systems that don’t crumble under thousands of tasks per minute. Here’s how the elite 2% do it 👇 ➤ Why monolithic agents fail—and how microservices save the day ➤ How Kubernetes orchestration powers self-healing, auto-scaling AI agents ➤ Load balancing tricks to prevent traffic spikes from taking you down ➤ Message queues that make workflows async, reliable, and resilient ➤ Why vector databases like Pinecone and Weaviate are the backbone of agent memory ➤ How observability + CI/CD pipelines transform fragile demos into self-improving systems Scaling AI agents isn’t a pilot project—it’s building production-grade reliability from the ground up. CC: Bhavishya Pandit

282

11 comentários

→ 𝐓𝐡𝐞 𝐅𝐮𝐭𝐮𝐫𝐞 𝐁𝐞𝐥𝐨𝐧𝐠𝐬 𝐭𝐨 𝐇𝐲𝐛𝐫𝐢𝐝𝐬 • AI is compressing single-skill value. Specialists alone can’t keep pace. • Cross-disciplinary problems arrive messy. Waiting on handoffs kills momentum. • One hybrid fluent in both domains outperforms two siloed specialists every time. → 𝐖𝐡𝐲 “𝐋𝐚𝐰𝐲𝐞𝐫 + 𝐃𝐚𝐭𝐚 𝐀𝐧𝐚𝐥𝐲𝐬𝐭” 𝐂𝐫𝐞𝐚𝐭𝐞𝐬 𝐚 𝐌𝐨𝐚𝐭 • AI Amplifies, Doesn’t Replace: AI can draft contracts or run models but cannot interpret nuanced intersections. • Integration Judgment: Hybrids reduce coordination overhead, translating constraints into opportunities faster. • Problem-Solving Fit: Real-world problems don’t fit in disciplinary boxes. → 𝐌𝐨𝐚𝐭 𝐃𝐫𝐢𝐯𝐞𝐫𝐬 • Rare & Valuable: Combining Law (10%) + Data (10%) yields <0.1% probability. • High-impact Areas: Litigation analytics, compliance, IP analysis, contract intelligence. • Example Value: A +1% win rate in major litigation can unlock millions; structured analysis of contracts makes thousands of documents actionable. → 𝐁𝐮𝐢𝐥𝐝𝐢𝐧𝐠 𝐘𝐨𝐮𝐫 𝐇𝐲𝐛𝐫𝐢𝐝 𝐒𝐭𝐚𝐜𝐤 • Depth First: 3–5 years mastery in primary field. • Avoid “Mediocre Twice”: Focus on complementary, high-synergy skills. • Deliberate Practice: 10–15 hours/week for 18–24 months. • Project-First Learning: One integrated project/month to solve real problems. • Market Integrated Value: Brand as “Legal Analytics,” not separate skills. • Golden 70/30 Rule: 70% primary, 30% complement. → 18-𝐌𝐨𝐧𝐭𝐡 𝐇𝐲𝐛𝐫𝐢𝐝 𝐁𝐮𝐢𝐥𝐝 𝐏𝐥𝐚𝐧 • Months 1–3: Assess strengths, shortlist complements, join communities. • Months 4–9: Skill build, complete integrated projects, publish use cases. • Months 10–15: Ship signature hybrid project, pilot paid engagements, apply for hybrid roles. • Months 16–18: Monetize skills, raise rates, plan next capability notch. → 𝐇𝐢𝐠𝐡-𝐕𝐚𝐥𝐮𝐞 𝐇𝐲𝐛𝐫𝐢𝐝 𝐑𝐨𝐥𝐞𝐬 • Clinical Analytics Lead: Healthcare + Data Scientist ($200K–$350K) • Logistics Systems Architect: Supply Chain + Software Engineer ($180K–$300K) • EdTech Product Lead: Teacher + Developer ($150K–$250K) • RevOps Strategist: Sales + Data Analyst ($180K–$280K) • AI Content Systems Architect: Writer + AI Engineer ($200K–$350K) → 𝐓𝐚𝐤𝐞𝐚𝐰𝐚𝐲 Hybrids win where single specialists fail. Rare skill intersections create moats, accelerate impact, and future-proof careers. Focus, deliberate practice, and integrated value creation are keys to building a lasting advantage. ---------------------------------------- Stop guessing what to learn next. Tell 𝐒𝐡𝐢𝐟𝐮 your goal → Get a personalized roadmap → Learn from curated content that actually moves you forward. Try your learning path on 𝐒𝐡𝐢𝐟𝐮. 👉 𝐉𝐨𝐢𝐧 the community to stay updated on new 𝐆𝐞𝐧𝐀𝐈-𝐀𝐠𝐞𝐧𝐭𝐢𝐜𝐀𝐈 advancements. Link in comments section 👉 𝐃𝐌 me for 𝐜𝐚𝐫𝐞𝐞𝐫 𝐠𝐮𝐢𝐝𝐚𝐧𝐜𝐞/ 𝐄𝐧𝐭𝐞𝐫𝐩𝐫𝐢𝐬𝐞 𝐀𝐈 𝐬𝐞𝐭 𝐮𝐩 Follow Ujjyaini Mitra for more insights on Enterprise Gen AI

45

18 comentários

OpenAI just launched AgentKit and a new Agent Builder, and it changes how we think about “AI tools.” Imagine building agents that help you research, draft, or automate simple workflows, without writing code. The real opportunity? Learning how to design them responsibly, with humans still in the loop. 🎥 Watch the 35-second video below, then tell me: How would you use an AI agent in your daily work? 👉 Tap follow for no-fluff AI news & insights from MIT’s Gen AI Global founder Oscar Sanchez 🧠 #AgentKit #AIAgents #FutureOfWork #NoCodeAI #GenAIGlobal

25

2 comentários

big improvements on our banking infrastructure for crypto companies. - we now support 5 chains (added solana, polygon, and eth mainnet) - added support for usdt - bring your own crypto wallet.

20

3 comentários

We’ve been rethinking what RAG really means at SCIKIQ. After building a few enterprise RAG systems, one thing became clear — most of them are still just search engines that summarize. They retrieve information, but they don’t understand it. If we really want AI that thinks and reasons like a human, we have to go much deeper. It starts with how we store knowledge. Not just text inside a vector database, but meaning — structure through Data Vaults, relationships through Knowledge Graphs, and semantics through Vector Stores. That’s when the system starts remembering context, not just matching text. Retrieval has to evolve too. It’s not about finding similar words, it’s about finding what’s most relevant, useful, and recent. Hybrid search, reranking, and context filters make that possible. Then comes reasoning — the real shift. Using frameworks like LangGraph or CrewAI, the system can plan, validate, and cross-check before it answers. It starts behaving less like a bot, and more like an analyst. At SCIKIQ, we’re combining this reasoning layer with controlled generation - so the output isn’t just fluent, it’s structured, explainable, and grounded in enterprise data. And over time, it learns. Every user correction, every feedback loop makes it a little sharper. That’s when automation starts turning into intelligence. Add trust and governance — lineage, versioning, and data source traceability — and now you have something enterprises can actually rely on. We call this next step beyond RAG a - Reasoning-Augmented Knowledge Fabric. AI that doesn’t just retrieve, but understands. Doesn’t just generate, but reasons. Doesn’t just automate, but grows with you. Maybe that’s where the future is heading — less retrieval, more reasoning. Less answers, more understanding. Less “system,” more thoughtful partner.

56

2 comentários

🚨 Is 'DeFi' the only use for Blockchain & Crypto?... No. Here's the short takeaway from our debate at Nova School of Business and Economics 💼 'Speculation is the use case' I have hear many say... 🤡 Or worse... 'Crypto is just a giant casino'... 🤯 Somewhere in the last 10 years, the level of thinking around the application of Blockchain technology has reduced to 'gamble on animal coins' rather than 'build something transformational'... 🥵 Yes, transformation is hard. But ultimately that when you create the most value... Leveraged gambling on magical internet money is NOT it... 🤷🏽‍♂️ So why is 'DeFi' not the only use case?... - Because tokenisation of 'things' was proven since 2017 - Because a shared record of encrypted data was proven since the 80s - Because ZK is transforming privacy in automated systems today - Because VeChain is rewarding 5 Million users worldwide everyday for completing sustainable actions with VeBetter dot com 😬 Saying DeFi is the ONLY use for Web3 is (intentionally from the organisers at Nova) an intellectual shortcut that may limit the imagination of those coming new to Web3 in 2025... 🔍 Study VeChain... Study VeBetter dot com, and the VeWorld 'Super App' and tell me if you still think Web3 is not suitable for everyday humans... What's your favourite non-DeFi use case for Blockchain technology? 🚀🌔 #blockchain #technology #management #innovation 

50

22 comentários

This is a quick one... I recently discussed with an exec how GenAI is transforming recommender systems, outperforming traditional methods. This is especially relevant for retailers with rapidly changing inventory, eg fashion brands regularly introducing new products. Yesterday, I found a quick and broad overview of these techniques that strikes a good level of detail. Sharing for anyone curious about these developments but looking for an accessible overview. https://lnkd.in/e93EtAXJ

6

⚙️ How fast can a 2B parameter LLM run natively on a MacBook Pro? Just benchmarked Microsoft’s BitNet 2B (W1.58A8, 4T tokens trained) locally via llama.cpp on Apple M2 Max and the numbers speak for themselves. 🧠 Model Load Time: ⏱️ 418 ms to load 1.1GB of weights, lightning quick for a 2B param model 💬 Prompt Eval Time: 🚀 9.7 ms/token for the prompt, or ~103 tokens/sec 🧾 Generation Speed: ⚡ 15.8 ms/token/ ~63 tokens/sec, fully local 📈 ~18.7K tokens/sec peak in the sampling chain 🧠 Memory Usage: 🧩 Metal GPU: 188MB 🧩 CPU: ~1.4GB total incl. KV cache + buffers 🔁 Architecture Efficiency: Offloaded 1 layer to GPU, the rest on CPU, smart hybrid execution using Metal (no CUDA or AVX) What does this mean? Running large language models on consumer hardware is not just possible, it’s practical. BitNet 2B delivers server-grade performance on a laptop, without compromising privacy or requiring cloud compute. This is the direction LLMs are heading: lightweight, accurate, fast, and local-first. Built with: 🛠️ llama.cpp 🧪 BitNet by Microsoft Keep pushing the boundaries of on-device AI. 💻🚀 #AI #BitNet #LLM #EdgeAI #Swift #JavaScript #AppleSilicon #llamacpp #opensource #localinference The AI Cowboys Roger Wong Won Glenn Brown Ed.D. PMP Corrina A. National Security Collaboration Center The University of Texas at San Antonio

22

1 comentário

Could alt szn be dead before it arrives? BTC just hit a new ATH, which is usually great for alts, BUT the money flow tells a different story; I know that sounds heretical on CT, but the data won’t lie just because it fit your bags. $45.6B flowed into stablecoins last quarter, a 324% surge from Q2. That’s not speculative money. That’s parked capital, patient, cautious, calculating. Meanwhile, VC funding into Web3 startups dropped 68% YoY. The signal is clear: liquidity isn’t leaving crypto, it’s just choosing a safer table. For years, the pattern was predictable, BTC pumps, ETH follows, alt szn funds the next wave of Web3 experiments. But this relationship is breaking.. This time, BTC’s rise isn’t dragging alts up with it. It’s detaching. Because the market no longer rewards culture, it rewards utility. And that brings me to a parallel worth studying. Satya Nadella, CEO of Microsoft recently laid off 10% of his workforce. The stock went up instead of plummeting. When asked why, he said: “The old equation = headcount equals value, doesn’t apply anymore. Efficiency does.” We’re seeing the same shift in crypto. The old relationship of — BTC rises = alts rally, doesn’t apply anymore. Efficiency does. Stablecoins and DeFi are the “Microsofts” of this cycle, efficient, essential, cash-flow driven. Speculative alts are still playing the old game, bigger culture, more tokens, more noise. So investors need to ask themselves: • Do altcoins hold real utility beyond being a pit stop for liquidity after Bitcoin runs? •Or are they still built on cultural momentum that fades when the charts flatten? Because the next bull run won’t reward vibes, it’ll reward fundamentals like; • Stablecoin rails. • Payments. • Composable infra. That’s where liquidity is hiding in plain sight. If you’re still betting on an automatic “alt szn,” you’re not investing, you’re praying. The correlation is broken. And in broken markets, the winners are those who stop waiting for the old cycle to return…and start building for the new one. So ask yourself: Are you still chasing hype, or positioning for where capital actually flows?

12

2 comentários

I recently sat down with Tomasz Tunguz of Theory Ventures to talk about agents that actually ship. Not a demo. A system that runs his firm. What we got into: • Why a few 𝘂𝗻𝗶𝗳𝗶𝗲𝗱 𝘁𝗼𝗼𝗹𝘀 beat a drawer of tiny ones • Exposing tools as a 𝗧𝘆𝗽𝗲𝗦𝗰𝗿𝗶𝗽𝘁 𝗔𝗣𝗜 instead of shell commands • Splitting 𝗽𝗹𝗮𝗻𝗻𝗶𝗻𝗴 from 𝗲𝘅𝗲𝗰𝘂𝘁𝗶𝗼𝗻 to keep agents predictable • Evals, traces, and observability as day-one requirements • The “email agent” that humbled both of us If you’re building with agents, this will save you a few faceplants. Full write-up is live on the Journey blog: https://lnkd.in/gaf7U4EE

18

1 comentário