Maybe it was taken out of context, but the headline ‘We don’t care about professional coders anymore’ is misleading and damaging to both non-coders and engineers. For non-coders: AI is incredible and improves daily, but it’s not fully independent. Think of today’s AI agents as

GPT-4.5 is out, and it’s a game-changer—but not how you think. My latest on Forbes forbes.com/sites/andrewfi…

There is probably no more controversial buzzword in software engineering right now than “vibe coding” — trusting AI to write the code for you. People love it, people hate it, they debate it on podcasts, and even teach online courses in it. Parents are questioning whether their

From embracing Agile early in my career to witnessing AI's radical transformation of software development today—⚡️ both revolutions follow the same pattern: helping teams create amazing products in a fraction of the time 👥 AI takes it further 🚀 by enabling "cross-functional

I'm looking forward to speaking about future-proofing the software development process at CTO Global Summit: event.geekle.us/cto?speaker #CTOGlobalSummit #Geekle

Just dropped Zen Agents! 🚀 No more solo AI coding assistants - we've built team-based AI for how devs actually work. Create custom agents, share them across your org, and automate those annoying in-between steps. 100+ tools ready in our open-source marketplace. Check it out:

After a decade in collaboration SaaS, I've seen firsthand how much time knowledge workers waste on reviews, approvals, and miscommunications. In fast-paced environments, deliverables spend most of their time "in between." AI can dramatically accelerate this process and flatten

Vibing about vibe coding youtube.com/watch?v=N9OHfC…

We have the next generational shift happening in agentic coding right in front of us: techradar.com/pro/the-three-… PS I wrote it about 6 weeks ago, and so far, everything has continued to accelerate in that direction.

🎙️ Just dropped: My conversation with pablosrugo on The Product Market Fit Show! We dive deep into: Bootstrapping Wrike to a $2B exit over 15+ years Why I didn't pivot (compound knowledge > starting over) Building Zencoder in the AI-first era Why velocity is everything in

Friends, if you’re juggling Cursor or Copilot and paying for ChatGPT Pro or Claude — check this out: Zencoder now lets you plug in your own AI subscription via their CLI tools. Time to find your zen. 👉 Sign up here: auth.zencoder.ai/signup 👉 Then configure OpenAI Codex or

🚀 Sonnet 4.5 is here — making advanced AI engineering accessible to everyone. We’ve been testing it ahead of release, and here’s what stands out: -Smarter at the same speed & price. You get Opus-level intelligence without Opus-level cost. -Remembers your work. Like collaborating

CEO of Zencoder, Andrew Filev, on Claude Sonnet 4.5: "It's very good at context management, which is one of the most important things to get these agents to do sizable tasks.” "They can just throw an idea at the model, and the model will automatically create the

Had a great conversation on The Information TV about what makes Sonnet 4.5 different. TL;DR - This model automates part of the work that expert AI engineers do today. It breaks down big tasks, writes specs, and manages context. It's similar to how reasoning models took

We might be witnessing the "iPhone moment" for AI – when the interface becomes the platform. New experiences, new companies, entire industries redefined: techcrunch.com/2025/10/06/ope… step for tech, giant leap for UX.

I remember the days when spam filters were terrible, and our inboxes were flooded with noise. Then Gmail came up with a decent filter and cleaned a massive chunk of that. Now, with AI SDRs in place, my inbox is total trash again. Unfortunately, the Gmail spam filtering team isn't

Haiku 4.5 is out. The model is remarkably capable — just six months ago, this level of performance would have been state-of-the-art on our internal benchmarks. Now it runs 4 times faster than Sonnet 4.5 at one third of the cost. This leap in performance and efficiency unlocks an

Andrew Ng has a point on evals that resonates with my current thinking: for complex agents, it's hard to build a 360-degree eval from scratch, but it is possible and easy to start capturing failure modes and build "regression evals" deeplearning.ai/the-batch/impr…