// Today in AI

Today in AI: Open Models Gain Ground Amid Security Concerns

The push toward open models continues as developers embrace alternatives like GLM 5.2, but enterprise caution emerges with Alibaba reportedly banning Claud

LDLatentDaily Desk Jul 4, 2026 1 min read

Today in AI — Photo by Markus Winkler on Pexels

July 04, 2026

The push toward open models continues as developers embrace alternatives like GLM 5.2, but enterprise caution emerges with Alibaba reportedly banning Claude Code over security risks. Meanwhile, clever cost-cutting hacks and new programming paradigms highlight the community's relentless innovation.

🚀 Developer ditches closed models for GLM 5.2 via Claude Code

A developer reports completely switching to open models, using GLM 5.2 through Claude Code daily. This signals a tangible shift in developer preference towards capable, open alternatives that offer more control and flexibility than proprietary options. (@_akhaliq)

💰 Alibaba to ban Claude Code over alleged backdoor risks

Alibaba is reportedly banning Claude Code in the workplace due to security concerns about potential backdoors. This shows major enterprises are taking a cautious, security-first approach when adopting third-party AI coding tools, potentially slowing widespread internal adoption. (Hacker News)

🔬 'Program-as-Weights' proposes new fuzzy function paradigm

New research introduces 'Program-as-Weights,' a programming paradigm for fuzzy functions that could change how we think about implementing uncertain or probabilistic logic in AI systems. This represents meaningful theoretical progress beyond just scaling existing architectures. (@_akhaliq)

🛠️ Clever hack cuts AI costs 60% using code-to-image conversion

Developers found a 60% cost reduction by converting code to images and having the model OCR it instead of processing code directly. This kind of ingenious workaround highlights both the high costs of current AI APIs and the community's creativity in optimizing them. (Hacker News)

🛠️ Guide to running state-of-the-art LLMs locally gains traction

A comprehensive guide for running cutting-edge language models locally is circulating, empowering developers to bypass cloud dependencies and costs. This reflects the growing maturity of local AI infrastructure and the desire for offline, private model operation. (Hacker News)

The takeaway: The tension between open model adoption and enterprise security concerns defines today's AI landscape, while cost optimization hacks show no sign of slowing down.