TLDR AI 2026-07-15
DeepSeek IPO plans 📈, Kalshi compute markets ⚡, Bonsai phone model 📱
Kalshi Ramps Up Effort to Build Markets for AI Computing Power (3 minute read)
Kalshi has created a tool that plots the future price of computing power so users can develop a sense of where the price to rent GPUs is headed. It will use prediction markets to build the forward curve. Forward curves are used to plot expected future interest rates and moves by central banks. Compute is becoming a commodity in its own right as demand grows. Kalshi's forward curve is based on weekly and monthly event contracts related to compute costs, and it goes as far as a year into the future.
Google announces Gemma 4 optimized for the Pixel 10's TPU (2 minute read)
Gemma 4 E2B for TPU is designed to run natively on the Pixel 10's TPU. It supports instant, deep offline conversation, image identification, and on-device audio transcription. The model also lets users command core phone functions, like Wi-Fi or maps, using just private voice or text. Video demonstrations of what the model can do are available in the article.
👨💻
Engineering & Research
Launch your website faster than ever with Framer. Now with Agents to bring speed and flexibility directly to your workflow. (Sponsor)
Framer is a professional website builder where AI agents make real, reviewable changes directly to your site. Trusted by teams at Miro and Perplexity, agents work across your canvas, components, CMS, and SEO to produce editable, production ready updates. Review every change in branches, compare diffs, and launch what works best.
👉 Launch your site with Framer today
WANDR Benchmark: Evaluating Research Agents That Must Search Wide and Deep (15 minute read)
WANDR (Wide ANd Deep Research) is an open benchmark and evaluation harness built around realistic and challenging data-collection tasks for knowledge work. Wide and deep research allows agents to sustain broad discovery without sacrificing factual quality from one record to the next. WANDR captures that structure with a flexible, composable qualification key hierarchy.
5 Trends That Defined AI Engineering at World's Fair 2026 (17 minute read)
AI engineering has come a long way in three years. The latest AI Engineer World's Fair shows just how much the field has matured. Engineering practices, such as building coding agents, designing harnesses, managing context, evaluating model outputs, and orchestrating increasingly autonomous systems, are now becoming part of mainstream software development. This post looks at the five largest trends that emerged during the AI Engineer World's Fair to see where AI engineering stands in 2026.
Announcing Bonsai 27B: The First 27B-Class Model to Run on a Phone (6 minute read)
PrismML announced Bonsai 27B, the first 27B-class model running on a phone, achieved through 1-bit and ternary weights, shrinking the model size to 3.9 GB and 5.9 GB respectively. Bonsai 27B supports complex tasks like multi-step reasoning and tool use while maintaining high performance, fitting securely within the device limits of modern smartphones.
The state of open source AI (15 minute read)
Open weights are now where the work happens. A majority of production tokens now route through them. The five highest-volume models on OpenRouter are all open. Closed models still lead at the frontier, but the frontier is not what most workloads need.
The Great Flattening (22 minute read)
Frontier coding models have shifted engineering's bottleneck from writing code to encoding judgment into agent harnesses that orchestrate planning, testing, review, and deployment. It is predicted that software organizations will flatten as cloud-based multi-agent systems replace traditional engineering workflows, leaving customer insight and product judgment as the primary human advantage.
Get the most interesting AI stories and breakthroughs delivered in a free daily email.
Join 1,100,000 readers for
one daily email