AI news · Programming & automation · 3 June 2026
DeepSeek V4 Flash became available free on the Nous Portal, making it possible to run a full Hermes Agent automation stack without paying for tokens. Julian Goldie tested the combination on real business tasks.
Read the summary →
AI news · Programming & automation · 25 May 2026
Google Antigravity 2.0 is Google's answer to Claude Code and OpenAI Codex — an AI-first development environment that plans, builds and tests applications autonomously. Julian Goldie ran a full application build from scratch. Here's what it produced.
Read the summary →
AI news · Programming & automation · 8 May 2026
Moonshot AI's Kimi K2.6 is a free open-source model built for coding. Julian Goldie paired it with Hermes Agent to build a dashboard frontend from scratch — here's what the session actually produced.
Read the summary →
AI news · Programming & automation · 7 May 2026
OpenAI Codex can take a keyword brief and produce a fully structured, publish-ready SEO page — headings, copy, schema, meta tags and HTML — in a single pass. Julian Goldie tested it on a real brief; here's what the output looked like.
Read the summary →
AI news · Programming & automation · 6 May 2026
OpenAI's Symphony is a free open-source spec that gives Codex-based AI agents a proper operating procedure — solving the problem of agents that start well but drift off-task halfway through complex jobs.
Read the summary →
AI news · Programming & automation · 2 May 2026
OpenAI's latest Codex update takes GPT-5.5 well beyond coding — it now controls apps, drafts and sends email, builds web pages, generates slides, and runs recurring scheduled tasks. Julian Goldie tested it. Here's what's genuinely new.
Read the summary →
AI news · Programming & automation · 29 Apr 2026
Xiaomi's MiMo V2.5 Pro scores close to paid models on coding benchmarks, runs free on the Nous Portal, and is already integrated into OpenClaw and Hermes Agent. Julian Goldie tested it against DeepSeek V4 — here's what held up.
Read the summary →
AI news · Programming & automation · 25 Apr 2026
OpenAI's new model scored 82.7% on Terminal Bench and can code autonomously for 20 hours — but the real capability only shows inside Codex, not ChatGPT. Here's what matters for a UK developer or small team.
Read the summary →
AI news · Programming & automation · 25 Apr 2026
DeepSeek V4 launches with open-source weights, free API access, and a 1M token context window. Julian Goldie tested it against GPT-5.5 and Claude Opus — strong on reasoning, less polished on design tasks.
Read the summary →