242 summarised stories in Business & Funding, each linking back to the original source. Browse all topics →
OpenAI Blog 9 months ago 33
OpenAI and AWS have formed a multi-year partnership to support OpenAI's AI infrastructure and model development. The deal is valued at $38 billion over its duration. AWS will supply the computing capacity needed to train and run OpenAI's next-generation models.
The daily briefing
TLDRocket reads 60+ sources, removes duplicate coverage, and summarises the day in two minutes. Free, no spam, unsubscribe anytime.
TLDRocket
3 August 2026
Onton has released Ontology 1, a neurosymbolic search model that outperforms Google Shopping and Amazon at the core task of e-commerce product discovery. The model scored a precision@10 of 0.630 versus Google's 0.543 and Amazon's 0.469 on a 90-query benchmark—a 16 percent improvement over the industry leader. Rather than relying on vector embeddings or seller-supplied labels, Ontology 1 reasons through an inspectable knowledge graph that maps product attributes, winning 52 of 90 test queries outright while indexing just 1 percent of competitor catalogs. The efficiency gain matters: the model does more with less, a constraint that matters in production systems where retrieval speed and catalog coverage both cost money. Onton is keeping the system proprietary, available only through live product deployment on Onton.com or case-by-case partnerships, which means adoption hinges on business deals rather than standard open or API access. That gatekeeping is deliberate—the company is betting that superior accuracy at a specific task (finding the right product for a given query) is defensible enough to build a business around. The deeper pattern here is the shift from pure neural search toward hybrid systems that combine learned representations with symbolic reasoning. Google and Amazon treat e-commerce search as a general ranking problem; Onton treats it as a knowledge representation problem, where explicit reasoning about product properties beats learned pattern-matching. It's a narrower bet, but precision matters when a user is looking for something specific.
Read the full briefing →