AI is getting more specialized—and the infrastructure bill is growing

A new comparison of image generators suggests there’s no single best choice for every task, with four models competing and several very recently released. For creators and product teams, that makes testing against a specific workflow more useful than picking by reputation alone.

iMerit has launched a platform to connect AI developers with specialists in particular fields. That could help teams bring relevant expertise into development and training, especially when general-purpose data or feedback isn’t enough.

Cognizant’s AI Nexus is presented as a modular, multi-model layer for overseeing enterprise AI spending and performance. For organizations using several models, the practical test will be whether a shared layer makes governance and cost decisions easier.

The initiative lets people contribute recordings, transcriptions and translations through an online platform. It’s a reminder that language resources can depend on community participation—and that local contributors have a role in how AI tools serve their languages.

Three Chinese labs have released coding-focused models within a week, with a reported 13-fold difference in input-token prices. Developers comparing them will need to weigh cost alongside how well each model handles their actual coding tasks.

FourKites is positioning its AI model around measurable supply-chain results rather than adoption of the software itself. For teams considering agents, that’s a useful bar to set: define the operational outcome before judging the tool.

A report says SpaceX raised $40 billion in debt financing to purchase GPUs for AI. For builders, it’s another sign that access to compute—and the capital required to secure it—can shape who is able to scale AI projects.

Nintex has introduced a process management product as part of a broader effort to link AI with workflows. The appeal for users is straightforward: AI is more useful when it fits into how work is already organised.

A Singapore project from BloodCounts aims to build AI models for analysing full blood count tests. It’s an early research initiative, and a reminder that healthcare AI needs careful validation before people can rely on it in practice.

A new piece asks whether data centres can support India’s AI leap. For the wider community, the question is not just how capable models are, but where the infrastructure to run them will come from.