Saw this question come up on my Linkedin, here is what I responded with: In the short term (next 5 years) AI will automate repetitive tasks employees are doing (tasks which SaaS is not doing) so no AI will not kill SaaS in the short term. Further lets use basic accounting system as example, asking AI to "Accurately" keep track of all transactions, do your taxes etc. will be massive waste of AI compute, SaaS accounting system is already good at doing just that, very accurately and very energy efficiently. Yes AI will be "smarter" than SaaS systems but the main constraint for AI is actual power usage, until AI can do all the work SaaS is doing, accurately at a lower energy point, it simply won't replace it. Will AI kill SaaS systems?

Just heard an enlightening conversation about the future of tech careers. Here's what you need to know: Get as close to the metal as possible Think about it: NASA won't use ChatGPT to launch rockets. The closer you are to core systems, the more valuable you become. It's like being a master mechanic in a world of self-driving cars – someone still needs to understand how things work at the fundamental level. Become an AI-Native Builder This isn't about coding anymore – it's about orchestrating AI to build entire systems. Think of yourself as a conductor, not just a musician. Key insights from the discussion: ☑️ The era of vendor lock-in is over – go open source, go local ☑️ Learning to communicate with AI agents is the new coding ☑️ Traditional development roadmaps are dead – adapt or fade ☑️Lean operations are crucial – use AI to amplify small teams Some memorable quotes: "The return on mastering core systems doubles every 6 months. While everyone chases the latest AI tools, fundamental skills become more valuable." "The ultimate test isn't how well you code, but how effectively you can communicate with AI to build solutions." 🔥 🤦‍♂️Back to basics #AITech #FutureOfWork #SoftwareDevelopment #TechCareers #OpenSource Thoughts? Let me know in the comments! 👇

🚀 Running DeepSeek R1 Distill Qwen &7b 8bit (mlx community) on my M1 Mac with 16gb memory - getting almost 23.139 tokens per second! Not bad for a 2-year-old Mac. No internet, No data leak risk, so total privacy. Plus there's no limit on the output, which is great. Planning to switch to using local code generation so that I can have unlimited generations. Anyone still paying for tokens? #DeepSeek #LocalAI #PrivacyFirst #UnlimitedAI #M1Performance

The AI landscape is shifting fast, and DeepSeek’s R1 has proven that smaller, more efficient models can challenge the might of OpenAI’s O1. But the real question isn’t just about size—it’s about who teaches models best! Welcome to Machine Learning 2.0 & Reinforcement Learning What's your take? What other skills do you think are essential for building effective LLM Models and AI agents?

This video and related X posts from Pliny the Liberator 🐉 caught my attention, since it 𝘀𝗵𝗼𝘄𝘀 𝗮 𝗱𝗶𝗳𝗳𝗲𝗿𝗲𝗻𝘁 𝘀𝗶𝗱𝗲 𝗼𝗳 𝘁𝗵𝗲 𝗔𝗜 𝗮𝗴𝗲𝗻𝘁 𝗰𝗼𝗶𝗻. ⚠️ 𝗗𝗼𝗻'𝘁 𝘁𝗿𝘆 𝘁𝗵𝗶𝘀 𝗮𝘁 𝗵𝗼𝗺𝗲. 𝗧𝗵𝗶𝘀 𝘄𝗮𝘀 𝗱𝗼𝗻𝗲 𝗳𝗼𝗿 𝗲𝗱𝘂𝗰𝗮𝘁𝗶𝗼𝗻𝗮𝗹 𝗽𝘂𝗿𝗽𝗼𝘀𝗲𝘀 𝗼𝗻𝗹𝘆. In a nutshell - Wes explores the implications of advanced AI systems, drawing attention to the actions of the notorious figure Pliny the Liberator aka @elder_plinius, who has been known for jailbreaking AI models. So even 𝗶𝗳 𝘁𝗵𝗶𝘀 𝘄𝗮𝘀𝗻'𝘁 𝗮𝗰𝘁𝘂𝗮𝗹𝗹𝘆 𝗲𝘅𝗲𝗰𝘂𝘁𝗲𝗱 - it shows more than implications of the potential that jailbroken LLMs have, as they are freed from their guardrails.