# AI News Digest - 2026-06-25

1. OpenAI and Broadcom unveiled "Jalapeño," a custom chip designed for large language model inference that was scheduled to run at scale by late 2026.

2. Anthropic released Claude Tag, which embedded the company's AI in Slack, and the company said the tool already generated 65 percent of internal product-team code.

3. Snowflake's CEO reported that Zhipu AI's GLM-5.2 nearly matched Anthropic's Opus 4.7 on a 103-task coding benchmark at about one-fifth the cost per output token while consuming nearly twice as many tokens per task.

4. Mistral AI released OCR 4 and said the model outperformed competitors in 72 percent of blind test cases for extracting text from documents.

5. MIT Technology Review published an analysis arguing that a new web data infrastructure layer was needed to provide real-time, trustworthy web data at scale to improve AI system performance and reduce hallucinations.

# References

1. [https://the-decoder.com/openai-and-broadcom-unveil-jalapeno-a-custom-chip-built-for-llm-inference/](https://the-decoder.com/openai-and-broadcom-unveil-jalapeno-a-custom-chip-built-for-llm-inference/)

[OpenAI and Broadcom unveil "Jalapeño," a custom chip built for LLM inference](https://the-decoder.com/openai-and-broadcom-unveil-jalapeno-a-custom-chip-built-for-llm-inference/)

1. [https://the-decoder.com/claude-tag-embeds-anthropics-ai-in-slack-already-writes-65-percent-of-internal-code-company-says/](https://the-decoder.com/claude-tag-embeds-anthropics-ai-in-slack-already-writes-65-percent-of-internal-code-company-says/)

[Claude Tag embeds Anthropic's AI in Slack, already writes 65 percent of internal code, company says](https://the-decoder.com/claude-tag-embeds-anthropics-ai-in-slack-already-writes-65-percent-of-internal-code-company-says/)

1. [https://the-decoder.com/snowflake-ceo-finds-glm-5-2-competitive-with-opus-4-7-at-a-fraction-of-the-cost/](https://the-decoder.com/snowflake-ceo-finds-glm-5-2-competitive-with-opus-4-7-at-a-fraction-of-the-cost/)

[Snowflake CEO finds GLM-5.2 competitive with Opus 4.7 at a fraction of the cost](https://the-decoder.com/snowflake-ceo-finds-glm-5-2-competitive-with-opus-4-7-at-a-fraction-of-the-cost/)

1. [https://the-decoder.com/mistrals-new-ocr-model-beats-competitors-in-72-percent-of-blind-test-cases-company-says/](https://the-decoder.com/mistrals-new-ocr-model-beats-competitors-in-72-percent-of-blind-test-cases-company-says/)

[Mistral's new OCR model beats competitors in 72 percent of blind test cases, company says](https://the-decoder.com/mistrals-new-ocr-model-beats-competitors-in-72-percent-of-blind-test-cases-company-says/)

1. [https://www.technologyreview.com/2026/06/24/1139202/the-emergence-of-the-web-data-infrastructure-layer-for-ai/](https://www.technologyreview.com/2026/06/24/1139202/the-emergence-of-the-web-data-infrastructure-layer-for-ai/)

[The emergence of the web data infrastructure layer for AI](https://www.technologyreview.com/2026/06/24/1139202/the-emergence-of-the-web-data-infrastructure-layer-for-ai/)

For the site tree, see the [root Markdown](https://ixtj.dev/.md).