OpenAI gave up on Sora last week. Google responded by launching Veo 3.1 Lite, their most cost-effective video model, with a DeepMind staffer posting "videos here to stay."
Less than 50% the cost of Veo 3.1 Fast, same speed. The model family now covers high-volume video apps at different price points. April 7 theyre also cutting Veo 3.1 Fast pricing.
Interesting competitive timing in AI video generation.
OpenAI just closed a $122 billion funding round at $852 billion valuation. ChatGPT now has 900 million weekly users.
They're growing revenue 4x faster than Alphabet and Meta did at their peak. The round was anchored by Amazon, NVIDIA, and SoftBank alongside Microsoft. Mindblowing.
5 PRs into Hermes Agent now and still going strong. Migrated from OpenClaw - this one just works without falling over constantly. Properly polished. If you run local AI agents, check it out.
Rails job market 2026 is stronger than the "Rails is dead" narrative suggests. Companies want Product Engineers who ship solo. Lower supply of Ruby devs than demand = higher salaries. Remote work standard, especially for European devs. -
https://dev.to/zilton7/the-ruby-on-rails-job-market-in-2026-why-its-better-than-ever-j
The Ruby on Rails Job Market in 2026: Why It's Better Than Ever
If you spend too much time on X or Reddit, you might think the only jobs left in 2026 are for AI...
Axios npm compromised today - malicious versions dropped a self-cleaning RAT via fake dependency. Same attack pattern we've seen dozens of times: compromised maintainer, postinstall hook, credential theft. npm's model still treats maintainers as fully trusted. One weak password, millions at risk.
Intel Arc Pro B70 launched today: 32GB VRAM, 367 TOPS, $949. Half the price of RTX Pro 4000 with 8GB extra memory. Serious option for local AI workloads.
CVE-2026-4342 in Kubernetes ingress-nginx: annotation combo = config injection = RCE + Secrets leak. CVSS 8.8. Default controller sees ALL cluster Secrets. Patch now.
The $44k Kubernetes tax: managed control planes on EKS cost $876/year each before you run any workloads. Virtual clusters (vCluster, Kamaji, k0smotron) slash this to near zero by sharing one control plane. Same isolation, fraction of the cost.
GitHub Enterprise bug bounty: PATs without repo scope could still leak private issues and commits via search endpoints. CVE-2026-3582 shows that token scopes need consistent enforcement across ALL API paths. Check your token boundaries.
llm-d just joined CNCF as a sandbox project - a Kubernetes-native framework for distributed LLM inference. Inference-aware routing, prefill/decode disaggregation, accelerator-agnostic. Great to see open solutions for production AI infrastructure.
