//devdigest

⚡️ DBSC lands in ASP.NET Core: experimental cookie-theft protection in .NET 11 preview 8 AddDeviceBoundSession() ties auth cookies to a TPM-backed key so a stolen cookie can't be replayed elsewhere; HTTPS-only, Chromium-only, and staying in preview past .NET 11 GA. ♻️ Subscribe for free now!

⚡️ IfItQuacks: duck typing in C# via interceptors, no dynamic or reflection IfItQuacks became a real library in part 2: a [DuckTyped] attribute plus a source generator check structural typing at compile time, now with events, indexers, delegates and static abstract members. ♻️ Subscribe for free now!

⚡️ The model price war: GPT-6 Luna is half the price of the last Luna, Opus 5.5 down 20% Claude Opus 5.5 and GPT-6 Sol/Luna landed with steep price cuts, but Opus 5.5 reportedly stalled at max reasoning on one test, burning its full output budget with no answer. ♻️ Subscribe for free now!

⚡️ Mark Seemann: LLMs speed up writing code, not how fast you actually learn Seemann argues LLMs don't shorten the real bottleneck in learning programming -- how fast a human absorbs concepts -- and suggests asking LLMs only falsifiable, verifiable questions. ♻️ Subscribe for free now!

⚡️ F# Weekly #40 is out: Blazor AI components for agent chat UIs Sergey Tihon's F# Weekly #40 is out. The highlight: experimental Blazor AI components in .NET 11 RC1 that turn an agent chat with streaming, tool calls and approvals into a single ChatPage component. ♻️ Subscribe for free now!

⚡️ Shopify compresses agent failures into model weights, cutting inference cost 96% A continual learning loop combining SFT, GRPO and prompt compression on PyTorch and vLLM let Shopify match frontier-model quality at a fraction of the cost. ♻️ Subscribe for free now!

⚡️ GPT-6 Astra, Sol and Luna land in Microsoft Foundry, starting at $0.10 per million tokens Three GPT-6 tiers for reasoning, general agent work, and high-volume routine tasks, priced $0.10-$75 per million tokens depending on model and direction, live across 28 regions. ♻️ Subscribe for free now!

⚡️ AG-UI Protocol gets an official .NET SDK built on Microsoft.Extensions.AI Microsoft shipped an official .NET SDK for the AG-UI agent-to-UI streaming protocol, built on IChatClient, interoperable with the existing TypeScript and Python SDKs. ♻️ Subscribe for free now!

⚡️ Transformers can now load llama.cpp GGUF quants directly via from_pretrained Hugging Face added native GGUF support for Qwen3.5 on Apple Silicon using ggml's Metal kernels, matching llama.cpp throughput while staying inside the regular Transformers API. ♻️ Subscribe for free now!

⚡️ Azure Container Apps Sandboxes: an isolated microVM for an agent in seconds Hardware-isolated microVMs with sub-second startup now run over a million agent sandboxes a day across Microsoft products, with KPMG running 30,000+ concurrent for tax work. ♻️ Subscribe for free now!

⚡️ Microsoft's live series builds a production AI agent in C# from scratch A four-part live series shows the Microsoft Agent Framework harness in C#: tool calls, approval policies, file sandboxing, shell execution and OpenTelemetry tracing, with code in the MafClaw repo. ♻️ Subscribe for free now!

⚡️ Rider and ReSharper 2026.2.2 add an AI Agent Setup widget and sharper coding-agent hooks Rider/ReSharper 2026.2.2 ships a status-bar widget for configuring MCP tools, skills and hooks, sharper inspection/reformat hook feedback for coding agents, and TUnit coverage support in dotCover. ♻️ Subscribe for free now!

⚡️ Environment.ProcessorCount reports process limits, not host cores, since .NET 6 Andrew Lock's platform-specific workarounds (Windows GetActiveProcessorCount, macOS sysctlbyname, Linux /sys/devices/system/cpu/online) recover the actual host processor count. ♻️ Subscribe for free now!

⚡️ Workflow1111 rebuilds the AUTOMATIC1111 feature set on a single gr.Workflow canvas Eleven media pipelines across 73 nodes, running without a GPU of its own. 22 of 36 operator nodes run in-process with no network call, every output node becomes a REST endpoint, and with mcp_server=True each one is an MCP tool. ♻️ Subscribe for free now!

⚡️ Azure Application Gateway adds HTTP/3 over QUIC in public preview Application Gateway now supports HTTP/3 over QUIC in preview, cutting connection setup time and avoiding TCP head-of-line blocking for mobile and API clients; HTTP/2 clients are unaffected. ♻️ Subscribe for free now!

⚡️ F# beats C# on most axes except tooling, argues Mark Seemann Seemann's core argument: F# disallows circular module dependencies by default, a language-level safeguard against spaghetti code that C# has no equivalent for. ♻️ Subscribe for free now!

⚡️ IBM's Granite PatchTST-FM-r2: SOTA zero-shot time series forecasting under Apache 2.0 The ~385M-parameter model ranks second for CRPS and MASE among replicable zero-shot models on GIFT-Eval and first among permissively licensed ones. Conformer blocks, 8,192-step context, 99 quantiles, Apache-2.0 and OpenMDW-1.0. ♻️ Subscribe for free now!

⚡️ Azure per-disk resiliency preview keeps a VM running when one data disk drops With per-disk resiliency, Azure takes only the affected data disk offline and reattaches it once connectivity returns, instead of recovering the whole VM. Announced alongside Azure Infrastructure Resiliency Manager. ♻️ Subscribe for free now!

⚡️ Uno Platform split their MCP server in two by data lifetime, not by feature A stateless docs server and a stateful local app server cost the agent 6.4k and 1.5k context tokens respectively, versus 5.2k for GitHub's single built-in MCP server. ♻️ Subscribe for free now!

⚡️ funes gives coding agents a memory built from their own session traces Hugging Face released funes, a local memory layer for Claude Code, Codex, pi and Hermes. It indexes session traces into a Lance dataset; recall returns original passages with provenance and can sync to a private Hub dataset. ♻️ Subscribe for free now!