#Tutorial/How-to

AIKarl's AI Watts asks whether a menu-bar tool can switch models while preserving the existing conversation, so users avoid re-explaining their project each time. The post frames this as the reason they want to keep the menu bar tool, which the quoted post describes as magpie, a menu-bar switcher for 20+ agents including Claude Code and Codex that also offers a local gateway.

AIHamel Husain reviewed Anthropic's new build_eval and hill-climb commands in the claude-api plugin for Claude Code, finding it useful for discovering issues like human handoff, formatting, and voice agent problems. He criticized it for pushing evaluator creation before data review, asking for label validation in Markdown files, and bundling four failure checks into one broad call-transfer evaluator. Husain says he would hold off on using it for now.

AIKling AI says Kling 4.0 will support native video generation from 3 to 30 seconds, with up to 10 keyframes for timing control. The full model is scheduled for October 2026, with early access to the new creation page rolling out from September 28.

AIGoogle Developers Blog describes how Sparse VideoGen (SVG) routes video diffusion attention heads into spatial or temporal sparse masks and implements them as custom JAX and Pallas Splash Attention kernels on TPU v6e. In isolated single-chip tests with 75.6K tokens and 10 heads, the sparse variants retain about 38.87% of query-key pairs. The article argues that theoretical sparsity must be converted into hardware tile skipping to yield real speedups.

AIGoogle Gemma says DiffusionGemma can be turned into a Jev-like model by seeding a canvas with a response template in vLLM. The approach extracts confidence scores and probability distributions for yes/no, multiple-choice, and scored questions in a single denoising step.

AIThe author argues that carefully crafted wording-level prompts have almost no effect in 2026, and that clear intent plus sufficient context matters most. Reusable prompt components are being absorbed into agent skills and context tools, while harnesses and models internalize more capability, leaving little room for prompting.

AISebastian Raschka traces text classification from bag-of-words models such as naive Bayes and logistic regression through pre-transformer neural networks, then sets up an analysis of the recently released Jev AI model. The article frames Jev as a general-purpose classifier that trades specialized accuracy for speed, cost, and breadth of tasks.

AIThomas Wolf, owner of the Hugging Face account, posted the single word "impressive" in response to a quoted post. The quoted post reports a new NanoGPT training record of 39.9s, down 27.7s from the prior 67.6s, achieved through per-flop optimizations such as sampled softmax and sparse updates.

AIThe article presents a four-part prompt structure (action verb, target element, desired result, protection instructions) for AI photo editing, saying it preserves approved work across platforms. It identifies three common failure causes: unmatched light direction, stacked edits in one prompt, and vague visual language. It states that simple skin retouching takes 2-3 minutes versus 15-30 minutes manually.

AIThe vLLM blog guide explains how separating prefill and decode, and moving tokenization to a CPU-only render tier, can keep token streams from stalling under load. In a two-L40S test on Qwen2.5-7B, collocated p99 inter-token latency reached 169 ms at 0.4 req/s while disaggregated serving stayed between 25 and 52 ms. The guide notes that the gain depends on fast KV cache transfer, and it includes setup code for NIXL-based serving and the render/derender API.

AISparse attention cuts per-operation KV cache reads but does not reduce overall memory capacity, so top-k cache misses still depend on HBM. SemiAnalysis's InferenceX estimates GB200 at about $0.044 per million total tokens at 150 tokens per second, roughly 12% below MI355X running ATOM at $0.049. Neither system holds a uniform cost advantage across the tested 100, 125, and 150 tokens-per-second targets.

AIPerplexity Developers released a cookbook recipe that combines Profiles, Skills, and managed connectors to build an incident responder. The recipe shows how the agent reads Datadog telemetry, diffs code in Sandbox, and posts an update to Slack.

AIGoodfire has published a practical guide to sparse autoencoders (SAEs), linking readers to the full blog post for details. The post itself offers little beyond the title and link, so no further specifics can be confirmed from it.

AIMastra's guide explains how teams can run agents under GDPR, with self-hosted deployments in any EU region or a platform environment created with --region eu. It covers PIIDetector redaction before data reaches the model, SensitiveDataFilter for trace fields, and retention and deletion handled in the team's own database. Mastra says it offers a DPA with EU Standard Contractual Clauses, a SOC 2 Type II audit, and no training on personal data.

AIMastra now offers classifiers that use evaluation models to answer questions defined as choice, score, or boolean, returning criteria keys, ordered positions, or true probabilities. Classifiers are registered on the Mastra instance and can drive workflow branching, such as routing a request to one of several agents. The feature requires @mastra/core 1.69.0 or later.

AIXiaomi MiMo reports that tool-call repetition in MiMo-V2.6 reached over 0.05% of responses across agent harnesses, causing stalled agents and wasted context. The team traced the cause to an RL flooding penalty set at 32 calls per turn, which missed smaller excess behavior, and replaced the approach with a specialized teacher distilled via MOPD. Repetition rates for both Pro and Flash dropped substantially, at roughly $90,000 versus an estimated $2.31 million for the alternative fix.

Why it matters: The post traces an agent failure to a reward blind spot and compares the costs of two fixes, offering a transferable debugging method for RL-trained tool-calling models.

AIXiaomi has released MiMo-V2.6-Pro-MOPD, an upgrade of the MiMo-V2.6-Pro-RL checkpoint that fuses several domain-specialized teachers into one model via MOPD2 and targets tool-call repetition. The sparse MoE model has 1.02T total and 42B activated parameters, a 1M-token context length, and accepts text, image, video, and audio inputs. Weights are available on Hugging Face and ModelScope, with deployment recipes for SGLang and vLLM.

AICanvases in the GitHub Copilot app are customizable interfaces that you and the agent share, such as kanban boards, dashboards, or checklists. You create one by running /create-canvas and describing the workflow, what you can do in the interface, and what the agent can do. Changes made by either you or the agent appear immediately in the shared canvas, and completed canvases can be saved as reusable extensions.

AIGoogle Cloud has launched a managed reinforcement learning fine-tuning service (RLFT) that lets customers adapt Gemini models using a reward function they define instead of labeled answers. Users supply prompts and a reward function, while Google handles the RL infrastructure and proprietary model internals. The guide advises exhausting prompting and supervised fine-tuning first, and notes that RLFT suits tasks that are easy to score but hard to demonstrate.

AIUsing the Neuron Kernel Interface, Reactor and Amazon's Neuron Science team built a kernel-centric path to real-time autoregressive diffusion video generation on Trainium. They addressed dynamic shapes, memory access patterns, and cache management, which are hard for generic compilers, and developed techniques intended to generalize across models.

AIAmazon Neuron Science and Reactor collaborated to optimize Rolling Forcing, an autoregressive diffusion video model, for real-time generation on AWS Trainium. The model generates 16 frames per second, the standard video playback rate, meeting latency requirements for streaming video generation.

AIvLLM has published a blog post explaining how its integration with TileRT works, alongside a public benchmark configuration in the InferenceX repository. The benchmark script covers GLM-5.3 FP8 on MI355X hardware and is linked on GitHub. The post itself provides the integration details.