← Back to archive
Radar just beat cameras at sea (and we got the receipts)

Image via Unknown

Tuesday, September 16, 2025

Radar just beat cameras at sea (and we got the receipts)

We're seeing some wild shifts in how AI systems actually work: radar's outperforming cameras on the high seas (yikes for vision-only approaches), OpenAI just dropped GPT-5-Codex for autonomous coding that sounds legitimately scary-powerful, and both OpenAI and Anthropic had their system prompts leaked online (oops). Meanwhile, Anthropic's new Global Economic Index is painting a concerning picture of uneven AI adoption creating economic divergence, while a clever new semantic ID approach is making LLM recommenders way more conversational and steerable without needing retrieval. But here's what's really getting to us: with these leaked system prompts and increasingly autonomous coding agents, are we building guardrails faster than the systems themselves?

Top Stories

1
Radar beats cameras in this sea test

Mythos AI's radar-based navigation system APAS is being tested aboard a commercial cargo ship as an alternative to camera-dependent approaches, designed to augment rather than replace human crew decision-making in maritime operations.

airoboticsmaritimeautonomous-systems
2
Semantic IDs for Recommender LLMs

Semantic IDs enable language models to natively understand and recommend from product catalogs by encoding items as meaningful token sequences, creating a unified search-recommendation-chat interface that combines the reasoning capabilities of LLMs with the behavioral prediction strength of traditional recommender systems.

llmrecommender-systemssemantic-idsfinetuning
3
System Prompt Leak

ThreadReader

Public system prompt leaks for OpenAI's GPT-5-Codex and Anthropic's Claude 3.7 Sonnet reveal detailed instructions for code generation, web integration, safety measures, and model capabilities. The leaks demonstrate how both companies engineer prompts to control AI behavior, support agentic tasks, and maintain accuracy across different use cases.

openaianthropicllmgpt-5
4
OpenAI Upgrades Codex with GPT-5

OpenAI

OpenAI's GPT-5-Codex brings autonomous coding capabilities directly into developer workflows with intelligent task complexity adaptation, code review automation, and seamless cloud-local integration. This positions AI-assisted development as a core part of enterprise software engineering infrastructure.

openaigpt-5codingagents
5
Anthropic Releases Global Economic Index Showing Uneven AI Adoption Patterns

Anthropic

Anthropic's Economic Index reveals deeply uneven AI adoption patterns tied to regional wealth and economic structure, with a notable shift toward autonomous task automation signaling growing user trust in AI capabilities. The findings raise important questions about economic inequality as AI adoption concentrates in wealthier nations and sectors.

anthropicclaudeai-adoptioneconomic-impact

Keep Reading

Industry Voices

Enjoyed this issue?

Get daily AI intel delivered to your inbox. No fluff, just the stories that matter.