Sponsored by
News of the day
1. IBM unveils Granite 4.2, a family of reasoning LLMs in 3B, 8B, and 30B sizes, featuring extended context windows and agentic capabilities. → Read more
2. OpenAI's custom Jalapeño chip boosts LLM inference performance, offering better efficiency and speed, with AI-generated code enhancing its capabilities. → Read more
3. Anthropic's new Playground tool outperforms OpenAI's established offering in key developer tests, excelling in export functionality and error handling. → Read more
4. Fossefall partners with Armada for AI factories, leveraging Luxembourg for funding and control of Nordic GPU infrastructure. → Read more
Our take
Hi Dotikers!
IBM has released Granite 4.2, and for once the interesting part isn't where it lands on the leaderboard.
Three dense models, 3B, 8B and 30B, Apache 2.0, pre-trained on roughly 15 trillion tokens, with a switchable thinking mode and a low-effort setting for questions that don't deserve the deliberation. The substance is in the post-training: the 8B and 30B go through an agentic reinforcement learning block where the model learns to actually act, fixing code in real repositories, driving a terminal, searching the web. The scores follow without startling anyone: 57 on SWE-Bench Verified for the 30B, 29 on Terminal-Bench 2.1.
The real product is the recipe. IBM lays out its curriculum stage by stage, its hyperparameters, its reward signals, what changes between the three sizes. This is the company your parents associated with mainframes publishing one of the most readable training write-ups of the year, while others ship a blog post and a contact form.
The other reason to pay attention goes back to yesterday. An agent tested by the UK AI Security Institute fabricated fake GitHub identities, staged a public apology and tried to get a malware dropper merged into an open source project. A student stopped it, not a security system. When an agent can lie to a maintainer for thirty-four hours straight, knowing where it runs and who can inspect its weights stops being a matter of principle. An open 30B model won't make your agent honest. It will simply go off the rails on your own infrastructure, in your own logs, without routing through an API you will never see inside of.
IBM's bet fits in one line: at some point, auditability will be worth more than three benchmark points.
Aym.
How AI-Era Pricing Is Reshaping Finance Operations
Usage-based and hybrid pricing models are changing how B2B companies generate revenue — and creating new headaches for the finance teams behind them.
Tabs co-founder Rebecca Schwartz and PwC Partner Amit Dhir sat down to unpack exactly what that means in practice: how pricing model decisions ripple into revenue recognition, forecasting, and financial ops — and what it takes to scale without piling on manual work.
Watch the on-demand recording to get practical frameworks, real-world examples, and a clear path to operationalizing usage-based revenue — including a forward-looking take on how AI will reshape financial workflows. If your team is navigating pricing complexity heading into the back half of the year, this is worth an hour.
Meme of the day
