Live AI developer news, ranked and linked to original sources.
> ▌

Ben Davis

AI Revolution

DIY Smart Code

DIY Smart Code

Theo - t3․gg

Eric Michaud

AI LABS

DIY Smart Code

Discover AI

Bijan Bowen

The PrimeTime

DIY Smart Code

Github Awesome

OpenAI

Income stream surfers

Two Minute Papers

AICodeKing

Prompt Engineering
Vercel AI Gateway now supports OpenAI's full-duplex GPT-Live 1.1 voice model, enabling real-time conversational streaming with per-second billing and no provider markup. Developers can integrate live audio via the AI SDK with WebSocket support and client-side model delegation for background tasks.
Cloudflare added Kuaishou's Kling Image 4.0 to its AI Gateway and Workers AI catalog, bringing 2K image synthesis and multi-image referencing to serverless edge applications. Developers can invoke the model with up to 10 reference assets at $0.063 per generation.
PlanetScale dogfooded its Insights Agent on its production Rails API, letting the AI analyze Query Insights telemetry and automatically open pull requests with database fixes. The internal rollout cut over 20 million daily queries and dropped latencies from seconds to milliseconds while uncovering crucial query-plan edge cases for agent evaluation.
Inworld has rolled out an optimized Gemma 4 26B serving pipeline delivering a 0.23-second p95 time to first token, up to 8.6× faster than the median provider on OpenRouter. The build lowers input pricing by nearly 40% to $0.07 per million tokens with context caching at $0.035, targeting latency-sensitive consumer and voice applications.
fal has enabled native 4K video generation for Kling 4.0 across text, image, reference, and video editing endpoints, introducing optional 10-bit HDR output at both 4K and 1080p. The upgrade allows developers to render high-fidelity video directly through the fal API starting at $0.549 per second for 4K output without relying on external upscalers.
Perplexity extended free access to its Decider-v1.3 and Flash models on OpenRouter for another three weeks as it reached nearly 20% of decision model requests. The specialized multimodal classifiers return calibrated probabilities and span extractions rather than text tokens to streamline agent routing, evaluation, and moderation.
OpenAI revealed Performance Factory, an internal agent orchestration system powered by Codex that autonomously monitors, investigates, and proposes code fixes for ChatGPT latency regressions. Across its initial rollout, the system tracked over 1,700 investigations and contributed dozens of production patches with human engineers in the loop.
Kong's Volcano platform combines isolated sandboxes, durable workflows, and branchable Postgres into a unified infrastructure stack tailored for autonomous coding agents. The system allows tools like Claude Code and Cursor to build, execute code safely, and deploy production services without fragmented cloud toolchains.
RunAnywhere's open-source TrnRT inference engine serves high-throughput large language models on AWS Trainium 2 instances using custom Neuron Kernel Interface megakernels and an OpenAI-compatible API. The architecture pairs two tensor-parallel engines per trn2.48xlarge instance with prefix caching and 1M-token context support for models like GLM-5.3-Flash.

Wes Roth

AI Search