🏥 In Practice

192 articles

🟡 🏥 In Practice July 10, 2026 · 2 min read

AWS: SageMaker Brings Serverless Fine-Tuning for NVIDIA Nemotron 3 Models with SFT, RLVR, and RLAIF Techniques

Amazon SageMaker AI has introduced serverless customization for NVIDIA Nemotron 3 models, requiring no infrastructure management. Three techniques are available: SFT (supervised fine-tuning), RLVR (reinforcement learning with verifiable rewards), and RLAIF (reinforcement learning from AI feedback), making advanced RL methods accessible to enterprise teams without ML infrastructure expertise.

🟡 🏥 In Practice July 10, 2026 · 2 min read

GitHub: Better Tools Made Copilot Code Review Worse — Rewriting Prompts Restored Quality at 20% Lower Cost

GitHub revealed that migrating Copilot code review to better-maintained tools initially made results worse — the cause was not the tools but outdated agent instructions. By rewriting the instructions to a diff-first approach (batch discovery before reading files, analysis anchored to the PR diff), they achieved around 20% lower average review cost with the same quality.

🟡 🏥 In Practice July 10, 2026 · 2 min read

OpenAI: Deutsche Telekom launches AI-native transformation — partnership covers support, network and voice AI for ~245 million customers

Deutsche Telekom and OpenAI have formed a partnership to transform the operator into an AI-native telco, with applications in customer support, employee tools, network operations and voice AI. Deutsche Telekom is one of Europe's largest operators with around 245 million customers globally, making this one of the biggest AI deals in telecommunications.

🟢 🏥 In Practice July 10, 2026 · 2 min read

Anthropic: Claude Code v2.1.206 Brings /cd with Path Suggestions, /doctor Advice for CLAUDE.md, and Auto Git Push in /commit-push-pr

Claude Code v2.1.206 is a new release of Anthropic's CLI tool that introduces a /cd command with path suggestions, a /doctor check that recommends trimming CLAUDE.md files, and automatic git push approval to the configured remote in /commit-push-pr. Background agents now update immediately after an upgrade, without a slow update on the next connection.

🟢 🏥 In Practice July 10, 2026 · 2 min read

AWS: Henry Schein One Verifies Dental X-Ray Quality with Real-Time AI — 11 Million Images per Week at 1.4-Second Latency

Henry Schein One has developed 'Image Verify', an AI system on Amazon SageMaker for real-time quality verification of dental X-ray images. The system has been deployed to more than 10,000 locations, processes over 11 million X-rays per week at an average latency of 1.4 seconds, with the goal of reducing insurance claim rejections caused by poor image quality.

🟡 🏥 In Practice July 9, 2026 · 2 min read

GitHub: Copilot now offers AI repository overview — automatic summary of purpose, technologies, and how to contribute

GitHub Copilot repository overview is a new feature that automatically generates a summary when first visiting an unfamiliar repository: what the project does, what technologies it uses, and how to contribute. It works on github.com through Copilot chat, is available on all Copilot plans, and requires no configuration.

🟢 🏥 In Practice July 9, 2026 · 2 min read

Anthropic: Reflect — dashboard showing how you use Claude, with quiet hours and self-reflection prompts

Reflect is Anthropic's beta dashboard for analyzing your own Claude usage habits: it shows dominant topics, task categories, and time patterns over periods of 1 to 12 months. Users can set quiet hours and break reminders, and it's available on Free, Pro, and Max plans with Memory enabled.

🟢 🏥 In Practice July 9, 2026 · 2 min read

Mistral: Studio gets version control for prompts and skills — immutable versions, audit trails, and rollback

Mistral Studio introduces a system of record for AI prompts and skills: immutable versioning with full history, ownership and audit trails, classification labels, and quick rollback. Skills are integrated as MCP servers, and development and production flows are separated — changes in production go through CI/CD.

🟡 🏥 In Practice July 8, 2026 · 4 min read

AWS Introduces Claude Apps Gateway: Centralized Management of Claude Tools for Enterprise Teams

Amazon Web Services introduced Claude Apps Gateway for AWS, a self-hosted control plane that gives enterprise teams centralized management of access, costs, and policies for Claude Code and Claude Desktop without distributing long-lived credentials to developers.

🟡 🏥 In Practice July 8, 2026 · 4 min read

OpenAI Challenges SWE-Bench Pro: The Leading AI Coding Benchmark Has Reliability Problems

OpenAI publishes an analysis questioning the reliability of SWE-Bench Pro — one of the dominant benchmarks for evaluating AI coding assistants in 2026. Since benchmark scores strongly influence purchasing and model adoption decisions, this warning has direct practical consequences for the industry.

🟢 🏥 In Practice July 8, 2026 · 4 min read

Claude Code v2.1.205: Session Security, Bug Fixes, and 400 MB RAM Savings

Anthropic released Claude Code v2.1.205 on July 8, 2026 with a new security rule that blocks manipulation of session transcript files, a series of fixes including silent JSON schema and Windows worktree bugs, and an auto-update optimization that now streams the binary to disk instead of into memory — saving approximately 400 MB of peak RAM.

🟢 🏥 In Practice July 8, 2026 · 3 min read

Flint — Microsoft's Open-Source Language That Compiles Visualizations for AI Agents

Microsoft Research released Flint, an open-source chart specification language that resolves the tradeoff between short and verbose visualization specs by compiling into Vega-Lite, Apache ECharts, and Chart.js from a single compact notation, with semantic data types that automatically derive scales, formatting, and color schemes.

🟢 🏥 In Practice July 7, 2026 · 4 min read

How Schneider Electric Built LLMOps Foundations for 60+ AI Agents with LangSmith

Schneider Electric, with 160,000 employees across 107 countries, deployed self-hosted LangSmith on AWS EKS to manage 60+ AI agents. The architecture rests on three pillars — observability, evaluation, and deployment — with approximately 200 active users. Internal assistant One Jo serves the entire organization, and a quotation workflow was reduced from days to 15 minutes.

🟡 🏥 In Practice July 6, 2026 · 4 min read

Alberta scanned 466 million lines of code in 20 hours: government uses Claude for cybersecurity

The government of the Canadian province of Alberta used Claude Code with approximately 50 parallel AI agents to scan 466 million lines of code in 20 hours — a task that would have taken 6.5 years using traditional methods. The project covered 1,280 applications and 3,400 repositories across 27 ministries, none of which had ever undergone a systematic security review.

🟢 🏥 In Practice July 6, 2026 · 4 min read

AWS automates PII redaction in images with Nova 2 Lite and SAM 3

Amazon Web Services has published a six-step pipeline for automatically removing personally identifiable information (PII) from images, combining Amazon Nova 2 Lite, Meta SAM 3, and Amazon Textract. The solution targets regulatory compliance with GDPR and PCI DSS without the need to train custom models.

🟡 🏥 In Practice July 3, 2026 · 4 min read

PASE: Neurosymbolic System Cuts Cloud Failure Recovery Time by More Than 40 Percent

Chinese researchers propose PASE — a Planning-Aware Semantic self-healing engine that combines LLM planning, symbolic verification, and deep RL prompt optimization. The result: more than 40 percent reduction in mean recovery time from cloud failures compared to prior approaches.

🟢 🏥 In Practice July 3, 2026 · 4 min read

How PyTorch Generates Thousands of Tests from a Few Template Classes — A Guide to the Test Infrastructure

A PyTorch blog post by Riya Punia (Red Hat) explains why CI failures carry strange names like TestMatmulCUDA.test_basic_cuda_float32 instead of the original TestMatmul.test_basic — tests are generated at import time through combinations of devices and dtypes.

🟡 🏥 In Practice July 2, 2026 · 4 min read

AWS SageMaker Guide: In Multi-Turn RL, Reward and Evaluation Matter More Than the Algorithm

The AWS SageMaker AI guide for multi-turn reinforcement learning places reward function quality and evaluation independence ahead of algorithmic choice. Dense rewards prevent variance collapse, and reward hacking occurs when the agent optimizes the metric without solving the actual task. On the SOP-Bench benchmark, correctly configured training achieved a 13 percent better Task Success Rate and around 16 percent better per-field accuracy.

🟢 🏥 In Practice July 2, 2026 · 4 min read

GitHub Copilot in CI/CD: No Personal Tokens and Team-Level Cost Controls

GitHub released two enterprise updates for Copilot on the same day: the CLI in GitHub Actions no longer requires a personal access token, and cost centers now support AI credit pools with automatic limits that protect teams from one another.

🟢 🏥 In Practice July 2, 2026 · 4 min read

How to Tame Exploding Coding Agent Costs: A Four-Phase Approach with LangSmith

LangChain describes how coding agents generate uncontrolled costs due to tool fragmentation and a tokenmaxxing mindset, and proposes a four-phase approach via the LangSmith platform covering visibility, normalization, optimization, and governance.

🟢 🏥 In Practice July 1, 2026 · 3 min read

Anthropic Introduces Model Access Control for Enterprise Administrators

Anthropic is launching enterprise model entitlements in beta: organization administrators can now specify which Claude models users have access to and configure effort-level settings per user or group.

🟡 🏥 In Practice June 30, 2026 · 4 min read

Claude Science: Anthropic's Dedicated Workbench for Life Sciences Researchers

Anthropic has released Claude Science in beta — a standalone desktop application (macOS/Linux) for researchers in genomics, proteomics, and related fields, with more than 60 curated skills, native integration of NVIDIA BioNeMo models, and a Reviewer agent that validates citations and calculations.

🟡 🏥 In Practice June 30, 2026 · 4 min read

OECD: GenAI Chatbot Usage Has Grown From 18% to 28% of the Population in One Year

An OECD analysis based on SimilarWeb data for 37 GPAI countries shows that the use of generative AI chatbots has grown from around 18% to 28% of the population between January 2025 and January 2026. Singapore leads with 63% adoption, the 25–34 age group surpasses 50%, and Claude and Copilot users gravitate toward professional tools.

🟡 🏥 In Practice June 29, 2026 · 2 min read

AWS: Nova 2 Lite and Claude — Cost-Optimized Pipeline for Processing Scanned Documents

AWS demonstrates a two-stage pipeline for mass digitization of scanned documents on Amazon Bedrock. Amazon Nova 2 Lite handles the initial extraction quickly and cheaply, while Claude takes on the spatially demanding understanding — a hybrid approach that costs two-thirds less than single-model alternatives.

🟡 🏥 In Practice June 28, 2026 · 2 min read

GitHub: Copilot agentic harness reaches vendor-harness level with lower token consumption across 20+ frontier models

GitHub Copilot agentic harness is a layer that gives models tools and an execution loop for autonomous coding. GitHub tested it on Claude Sonnet 4.6, Claude Opus 4.7, GPT-5.4, and GPT-5.5 across five benchmarks and found that it achieves task completion rates on par with vendor harnesses while consuming fewer tokens, with support for more than 20 frontier models.

🟡 🏥 In Practice June 27, 2026 · 2 min read

Anthropic: Economic Index reveals daily rhythms and patterns of AI usage

The third Anthropic Economic Index report analyzes when and how people use Claude — from morning news to late-night recipes. Claude Code's autonomy score is 0.37 points higher than chat, and conversations in high-paying occupations consume 2.5x more tokens.

🟡 🏥 In Practice June 27, 2026 · 2 min read

arXiv:2606.27009: Semantic early stopping cuts agent loop token cost by 38%

Semantic Early-Stopping for Iterative LLM Agent Loops proposes a method that halts an agent's iteration loop as soon as the embedding of successive drafts stops changing semantically — without a fixed step count — reducing token consumption by 38% with identical quality.

🟡 🏥 In Practice June 26, 2026 · 2 min read

arXiv:2606.25519: Quantization inflates reasoning — the hidden cost of low-bit models

Quantizing language models to INT4/INT3 preserves answer accuracy but lengthens the chain of thought and cancels the expected inference speedup. Microsoft researchers introduced the CoT Token Inflation Ratio metric and tested it on math, code, science, and agentic tasks.

🟢 🏥 In Practice June 26, 2026 · 2 min read

arXiv:2606.25524: Cliff Tokens — single tokens that trigger failure in mathematical reasoning

Cliff tokens are individual tokens in an LLM output where the probability of successful mathematical reasoning drops sharply. Researchers developed a detection method and showed that removing the first cliff token restores accuracy to near-perfect levels, while Cliff-DPO training yields +6.6 percentage points.

🟢 🏥 In Practice June 26, 2026 · 2 min read

Microsoft: Generative Causal Testing — AI hypotheses about the brain validated by scanner

Generative Causal Testing (GCT) is a two-phase AI framework that converts opaque brain activity prediction models into testable hypotheses — and then verifies them with real fMRI experiments on human subjects.

🟡 🏥 In Practice June 25, 2026 · 2 min read

AWS: Huntington Bank redacted PII from 400 million documents at 95% accuracy

Huntington Bank used an AWS stack (Textract, SageMaker, Step Functions) to redact personal data from more than 400 million documents at over 95% accuracy, cutting the original cost estimate to just 5% and compressing timelines from years to months.

🟡 🏥 In Practice June 25, 2026 · 2 min read

Microsoft: Talos — open-source automated iterative genomic reanalysis for rare diseases

Talos is an open-source system that automatically and iteratively reanalyzes genomes of patients with rare diseases. Applied to 4,735 undiagnosed patients, it found 241 new diagnoses (+5.1%), with 90% sensitivity and an annotation cost of just $11 per 1,000 genomes.

🟡 🏥 In Practice June 24, 2026 · 2 min read

AWS: Bedrock AgentCore pool-model multi-tenancy — shared infrastructure, isolated tenants

AWS Bedrock AgentCore introduces a pool-model multi-tenancy architecture with three-tier isolation (Tier → Tenant → User), Cedar policies for tool boundaries, and a Token Vending Machine for memory isolation — a reference SaaS design for production AI agents.

🟢 🏥 In Practice June 24, 2026 · 2 min read

OpenAI: GPT-5 Pro helped solve a 3-year immunology mystery about T-cells

GPT-5 Pro helped immunologist Derya Unutmaz of the Jackson Laboratory resolve a three-year scientific mystery about T-cell behavior — immune cells critical to the body's defense — opening new possibilities in cancer and autoimmune disease research.

🟢 🏥 In Practice June 23, 2026 · 2 min read

AWS: semantic search of aerial imagery with Amazon Nova Multimodal Embeddings (Vexcel)

Vexcel and AWS demonstrated semantic search of aerial photography using Amazon Nova Multimodal Embeddings. After testing around 100 configurations, LLM-generated captions improved the F1 score for swimming pools by 11% and for roads by 13%, which evolved into the commercial product Vexcel Intelligence available in 45+ countries.

🟢 🏥 In Practice June 23, 2026 · 2 min read

CNCF: why classical observability does not work for agentic and LLM systems

The CNCF blog explains why the classical approach to system monitoring does not work for AI agents and LLM models that operate probabilistically — the same query can produce entirely different results, and errors are semantic rather than technical.

🟢 🏥 In Practice June 21, 2026 · 2 min read

arXiv:2606.20474: UltraQuant Reduces KV Cache Latency by 3.47× with 4-Bit Precision

UltraQuant is a KV cache compression technique targeting 4-bit precision for multi-turn LLM agents. Developed at AMD, UCLA, and Purdue, it achieves 3.47× faster P50 TTFT in late rounds under high context pressure and 1.63× higher output throughput compared to the FP8 baseline.

🟡 🏥 In Practice June 20, 2026 · 2 min read

Anthropic: Claude Code v2.1.183 Blocks Destructive Git and Infrastructure Commands in Auto Mode

Claude Code v2.1.183 is the new version of Anthropic's CLI tool released on June 19, 2026, which in auto mode blocks destructive git commands (git reset --hard, git clean -fd, git stash drop) and infrastructure commands (terraform, pulumi, and cdk destroy) without explicit user request. The version also brings fixes for WebSearch in subagents, fullscreen UI corruption, and MCP authentication.

🟡 🏥 In Practice June 20, 2026 · 2 min read

AWS: SageMaker Gets Over 100 Detailed Inference Metrics and an Insights Dashboard on CloudWatch

Amazon SageMaker is AWS's machine learning platform that now emits more than 100 detailed metrics for generative AI inference into CloudWatch, available through the new Insights dashboard. The dashboard displays token-level latency (TTFT, inter-token latency), KV cache tracking and cold start diagnostics, and is compatible with Grafana and Datadog via a PromQL endpoint.

🟢 🏥 In Practice June 20, 2026 · 1 min read

GitHub: Copilot Retires Opus 4.6 (fast) on June 29, Adds AGENTS.md to Code Review and ai_credits_used Field to API

GitHub announced three changes for Copilot through its changelog. The Opus 4.6 (fast) model is being retired on June 29, 2026 across all Copilot surfaces with a recommendation to switch to Opus 4.8 (fast), Copilot code review now reads repository AGENTS.md files, and the Copilot Usage Metrics API gains the ai_credits_used field tracking AI credit consumption per user.

🟡 🏥 In Practice June 19, 2026 · 2 min read

arXiv:2606.19245: TxBench-PP — AI Agents in the Search for New Drugs

TxBench-PP is a benchmark testing AI agents in preclinical small-molecule pharmacology across 4,800 trajectories and 11 models. Claude Opus 4.8 achieves 59.3% success and leads ahead of GPT-5.5 at 55.3%, but no model reaches the reliability needed for medical application.

🟡 🏥 In Practice June 19, 2026 · 2 min read

GitHub: MAI-Code-1-Flash Available on 8 Development Surfaces Within Copilot

MAI-Code-1-Flash, Microsoft's compact coding model positioned as best-in-class for its size, is now available on 8 development surfaces within GitHub Copilot — from CLI to mobile platforms — across all plans from free to max tier.

🟡 🏥 In Practice June 19, 2026 · 2 min read

OpenAI: AI Helps Physicians Diagnose Rare Genetic Diseases in Children

OpenAI's reasoning model identified 18 new diagnoses in previously unsolved cases of rare genetic diseases in children, collaborating with physicians and research institutions. The results open new questions about the role of AI in clinical diagnostics.

🟢 🏥 In Practice June 19, 2026 · 2 min read

PyTorch: LLMs Reduce GPU Kernel Optimization from Minutes to Seconds

The PyTorch core team published LLM-guided autotuning for Helion kernels that accelerates GPU code optimization from minutes to seconds. Instead of exhaustive search across all configurations, large language models intelligently guide the parameter space search.

🟡 🏥 In Practice June 18, 2026 · 2 min read

GitHub: Copilot desktop app now generally available — parallel sessions and cloud automations

GitHub has announced that the Copilot desktop application for macOS, Windows, and Linux is now generally available. The app brings parallel sessions, a canvas interface, cloud automations, and integration of custom models and tools. It unifies GitHub Spark, Copilot Chat, and Copilot CLI into a single desktop experience. The release is part of a broader wave of Copilot announcements the same day, including HyDRA model routing and Auto mode available to all users.

🟡 🏥 In Practice June 18, 2026 · 2 min read

IBM: Study Reveals 91% of Organizations Lack Full Visibility into AI Dependencies, Lock-In Serious

An IBM study shows that 91% of organizations lack full visibility into their own AI dependencies on vendors, models, and infrastructure. As many as 71% would struggle to change their primary AI vendor, while 81% say a seven-day vendor outage would cause severe or critical disruption. Only 7% have advanced AI control capabilities, yet those organizations protect operating profit from disruptions 55% better. As many as 72% would pay 20% more for vendor flexibility.

🟢 🏥 In Practice June 18, 2026 · 1 min read

Anthropic: Claude Code v2.1.181 introduces /config syntax and Bun 1.4 with auto-retry on API failures

Anthropic has released Claude Code v2.1.181 with a new /config key=value syntax for setting configuration directly from the prompt and an upgrade of the Bun runtime to version 1.4. The release adds automatic retry on API connection drops during the 'thinking' phase, hides inactive subagents after 30 seconds, and speeds up startup by 120 ms in fresh environments. Prompt caching with a custom ANTHROPIC_BASE_URL and Foundry has also been fixed.

🟡 🏥 In Practice June 17, 2026 · 2 min read

Anthropic: Agentic Coding Study — Users Make 70% of Planning Decisions, Claude 80% of Execution Decisions

Anthropic published an economic study of agentic coding with Claude Code. Users make approximately 70% of planning decisions, while Claude takes approximately 80% of execution decisions. Expert users receive 12 actions and 3,200 words per prompt versus 5 actions and 600 words for beginners, with verified task success rates of 28–33% versus 15%. The share of debugging sessions fell from 33% to 19% between October 2025 and April 2026, while the average task value rose approximately 25%.

🟡 🏥 In Practice June 17, 2026 · 1 min read

AWS: Container Caching in SageMaker AI Cuts Inference Scaling Latency by Up to 50%

AWS introduced container image caching in Amazon SageMaker AI that eliminates ECR pulls during scaling, automatically and without any opt-in required. For Qwen3-8B, startup latency dropped from 525 to 258 seconds, approximately 51%. Users report 38–65% lower P50 latency and up to 2× faster end-to-end scaling. The feature is available on all accelerator instance types across all commercial AWS regions.

🟡 🏥 In Practice June 17, 2026 · 1 min read

Google DeepMind: AI Tool for UK Planning Targets 50% Faster Decisions and 1.5 Million New Homes

Google DeepMind, the UK government, Google Cloud, and Faculty AI are developing an AI prototype targeting 50% shorter decision times for planning applications. The government's goal is 1.5 million new homes by 2029. Householder applications account for nearly 70% of all annual applications, with an estimated saving of approximately 255 hours per year per council. The tool is currently being tested in Barnet, Camden, and Dorset, with planned nationwide rollout in 2027.