AGI API

One gateway to access every frontier model.
A unified AI gateway for developers, professionals, and early adopters.

[AGI API @agiapiorg]

MIT Technology Review reports that Texas is the fastest-growing US data center market and ground zero for Big Tech's largest AI buildouts, but the blue wave that swept the state may jeopardize that position.

[AGI API @agiapiorg]

The Guardian reports Elon Musk poured $446m into Republican-aligned PACs for the midterms, yet Democrats won all five Senate races where his America Pac spent the most.

[AGI API @agiapiorg]

OpenAI outlined a standard for responsible AI infrastructure, arguing that building more US computing capacity can lower the cost of delivering AI and put the tools in more people's hands.

[AGI API @agiapiorg]

The Decoder reports that after US voters rejected new data centers in the midterms, OpenAI is launching a charm offensive with voluntary commitments that have gaps.

[AGI API @agiapiorg]

Together AI released Together Compiler (TKCC), an agent-driven system that keeps re-optimizing inference as production traffic changes. It already runs on several serverless endpoints and can make inference up to 2.6x faster.

[AGI API @agiapiorg]

Kairos, an open-source risk and fraud decisioning platform on GitHub, makes real-time decisions on sign-ups, logins, orders, payments and payouts and keeps the signals behind each one.

[AGI API @agiapiorg]

Anthropic's Frontier Red Team released FlashProve, an open-source zkSNARK prototype that proves a reply came from a committed model without revealing weights. It hits about 110 tokens/sec on Llama 4 Maverick with one H100.

[AGI API @agiapiorg]

The Guardian reports US tech stocks recovered after a selloff over AI spending fears, with midterm elections seen as unlikely to have a huge impact on Wall Street.

[AGI API @agiapiorg]

Fast Company reports that a new Texas governor could change the state's AI boom, less than a year after Greg Abbott celebrated Google's $40 billion investment.

[AGI API @agiapiorg]

MarkTechPost reports Soniox released STT v6, a speech-to-text API that uses the voice agent's text to recognize short replies and filter the agent's own voice; streaming and file versions cover 60+ languages.

[AGI API @agiapiorg]

NVIDIA's developer blog explains automated workload recovery for distributed AI training, covering how NCCL timeouts, unresponsive GPU nodes and stragglers stall jobs and why recovery overhead matters for goodput.

[AGI API @agiapiorg]

Zyphra released Zyphra Hierarchical Attention, a sparse attention method with a tree-structured indexer. It distills into frozen pretrained models with about 4B tokens and stays close to dense attention on Qwen3.6-35B-A3B benchmarks.

[AGI API @agiapiorg]

TechCrunch reports Apple's AI-powered Audio Intelligence for Apple Watch is now in beta testing for developers in iOS 27.3, with four opt-in features for Series 12 including conversation recaps and sound detection.

[AGI API @agiapiorg]

The Letter Company released the beta of Letterpress-1, a coding agent for growth engineering that reads a site and its market, builds changes in a sandbox and opens a pull request with a live preview.

[AGI API @agiapiorg]

Tech Policy Press argues the SEC should block Anthropic's IPO, pointing to circular AI deals and murky finances that obscure an accurate financial picture for investors and regulators.

[AGI API @agiapiorg]

Simon Willison says that since switching to Python 3.15 he has hit occasional confusing dependency errors, and the rule he keeps having to remember is: if in doubt, delete the uv.lock file.

[AGI API @agiapiorg]

AI Business reports OpenAI delayed GPT-6.1 Astra after testing exposed new safety concerns, highlighting why models need continued evaluation and enterprise runtime controls.

[AGI API @agiapiorg]

Measure the carbon cost of everyday items: prompts, beef, flights.

[AGI API @agiapiorg]

Hacker News is discussing which TV to recommend for privacy, including which sets pair best with Apple TV for a display-only setup.

[AGI API @agiapiorg]

Fast Company auditioned 12 AI agents for a chief-of-staff role, comparing how they manage inboxes, book travel, organize files and catch mistakes.