OpenAI's Unification Play: One App, Three Products, No Fanfare
OpenAI quietly announced it's merging ChatGPT, Codex, and Atlas into a single desktop app. The April Codex update shows what unification looks like in practice.
briefings
All briefings
Reverse chronological. 12 total.
The frontier becomes a plateau: three models tie at 57
Claude Opus 4.7, Gemini 3.1 Pro, and GPT-5.4 all score 57 on Artificial Analysis—the first three-way tie at the top in eighteen months. Intelligence parity masks stark differences in price and speed.
What We Actually Know About Grok's Architecture
The only publicly documented Grok architecture is Mixture of 8 Experts — not 16 agents. Meanwhile, xAI's real story is nine departed co-founders and a paused agent project.
DeepSeek V3.2: The 31x Price Gap That Changes the Math
DeepSeek's V3.2 costs $0.32 per million tokens while scoring within two points of Claude Sonnet 4.6. At 31x cheaper than Opus, the math starts mattering.
The naming wars: Clawdbot, Moltbot, OpenClaw
An open-source AI assistant framework went through three names in a month after its Claude-adjacent original triggered trademark concerns from Anthropic.
Anthropic's Cross-Platform Play: What Actually Ships Today
While competitors focus on chatbots and APIs, Anthropic has quietly built a multi-surface product line: Claude Code across five platforms, mobile apps with 109K ratings, and computer use that lets Claude operate your Mac.
Google's AI Coding Play: Plugins and CLI, Not a New IDE
While Antigravity remains unverifiable vapor, Google's actual AI-coding strategy is distributed: Gemini Code Assist plugins with agent mode in preview, and the 102k-starred open-source Gemini CLI.
Anthropic ships Opus 4.7, and nobody quite understands what 'xhigh' means yet
Anthropic releases Opus 4.7 with adaptive thinking, 128k max output, and breaking API changes—all at unchanged pricing. Customer benchmarks show 12-20% gains.
Horowitz says software's laws of physics have changed
At a16z's Connect/Fintech conference on April 14, 2026, Ben Horowitz argued that AI has rewritten the rules under software — collapsing moats, compressing product runway to weeks, and making US infrastructure the real constraint.
Four agents walk into a browser
Google, OpenAI, Anthropic, and Perplexity have all shipped browser agents. The form factor converged in eighteen months. The benchmarks haven't.
Perplexity Computer puts 19 models to work — if it works
Perplexity shipped a cloud agent that orchestrates 19 AI models. Then they canceled their own demo hours before the press briefing.
The $800 Lesson in Vibe Coding
A YouTuber's surprise Vercel bill reveals the hidden costs of letting AI agents choose your infrastructure—and the deeper question of whether humans can still review the code they're shipping.