Signal

Loading the stream…

September 2026

Score 4.14curatedAgent tools & setup
hamel.dev (Hamel Husain) · T1.52026-09-30original ↗ · G·2500
Claude’s new auto eval tool
Summary
Hamel Husain and Isaac Flath tested Anthropic's new Claude Code eval tooling (build_eval, hill-climb) on apartment-leasing conversation traces. They identified concrete problems: it suggests evals before users review data, and asks for judgment validation without sufficient context for inline annotation.
Why it matters
First-party Anthropic tooling is likely to be widely adopted; a hands-on critique from a recognized practitioner flags real workflow pitfalls and suggests better alternatives.
Score 4.40curatedAgentic workflow patterns
arXiv cs.MA · T12026-09-30original ↗ · G·2487
Prompted Identity Degrades Cooperation in Multi-Agent LLM Systems
Summary
Empirical study showing that when multi-agent LLM systems are aware of each other's model family, they form identity-based clusters that impair cooperation. Labeled groups use 30% more rounds and 55% more tokens, with success rates dropping from 96% to 81%; withholding identity labels eliminates the effect.
Why it matters
Controlled experiment isolating a previously unmeasured failure mode in mixed-provider multi-agent setups, with a single-line mitigation practitioners can apply to their own systems.
Score 3.80curatedAgentic workflow patterns
arXiv cs.MA · T12026-09-25original ↗ · G·2379
AI-Moderated Interviews for Market Research and Digital Twins Calibration
Summary
A pre-registered study (N=317) with three industry partners compares AI-moderated, human-moderated, and static interviews for market research. AI moderation matches human depth and covers more themes, but the extra richness does not improve consumer digital twin predictions over static interviews; prediction errors track thinking-style gaps and out-of-distribution questions.
Why it matters
Empirically tests whether AI-conducted interview depth translates into better digital twin predictions. The negative result is a useful corrective for teams building or buying such pipelines.
Score 3.80curatedAgent tools & setup
arXiv cs.HC · T12026-09-24original ↗ · G·2332
Safety Nudges: User-Facing Interventions for Real-Time AI Risk Awareness
Summary
Paper introduces Safety Nudges, a browser tool that flags concerning chatbot behavior (hallucination, sycophancy, overconfidence, anthropomorphism) in real time. A two-week field study with 45 frequent chatbot users found the tool useful, clear, and minimally disruptive, and increased risk awareness, but awareness did not translate into measurable behavioral change.
Why it matters
Field-tested design with a counter-intuitive finding: surfacing AI risks to users raises awareness without changing behavior, so calibration and user control matter more than alert volume.
Score 3.80curatedAgent tools & setup
GitHub Changelog · T12026-09-23original ↗ · G·2310
Local sandboxing in the GitHub Copilot app
Summary
GitHub Copilot app now supports local sandboxing, restricting the agent's access to files, network resources, and credentials on the user's machine to reduce the impact of unintended commands.
Why it matters
First-party GitHub feature that adds a security boundary for agent-executed commands. Worth configuring for anyone running Copilot coding agent on real work.
Score 4.20curatedPapers & ergonomics
arXiv cs.HC · T12026-09-22original ↗ · G·2277
Adaptive Scaffolding Needs Contingency: An AI Tutor That Escalates and Fades on What the Learner Does
Summary
Within-subjects study with 131 learners tested CoMeT, an AI coding tutor built on 'preserved metacognitive demand,' which holds learner decision-making constant while varying support. CoMeT escalated help on failure and faded on uptake, matched a question-only tutor's demand, produced twice the artifacts of an unrestricted assistant, and frustrated learners less.
Why it matters
Distinguishes cognitive load from metacognitive demand as a design axis for AI tutors. Controlled comparison, specific escalation/fading rules, and quantified surrender rates make it directly applicable.
Score 3.80curatedAgent tools & setup
GitHub Changelog · T12026-09-14original ↗ · G·2133
Configure cost and quality in Copilot auto model selection
Summary
GitHub Copilot's auto model selection now offers three configurable tiers — efficiency, balance, and intelligence — letting users set how the system weighs cost, quality, and response time.
Why it matters
First-party release of a concrete Copilot configuration knob; anyone using auto model selection should know how the three tiers map to cost and quality tradeoffs.
Score 3.78curatedAgentic workflow patterns
Addy Osmani · T1.52026-09-14original ↗ · G·2251
Brownfield Agentic Engineering
Summary
Addy Osmani's piece on Brownfield Agentic Engineering examines the requirements and challenges of running AI agents in existing legacy codebases, where the code predates the current team working on it.
Why it matters
Most agent discourse covers greenfield demos; this tackles legacy codebases where most real engineering actually happens.
Score 3.80curatedAgentic workflow patterns
arXiv cs.HC · T12026-09-10original ↗ · G·2025
Where Does the Human End? Creative Agency with Generative AI across Five Years of Chinese Digital Painting
Summary
A five-year interview study (2021–2025) with 17 Chinese digital painters tracks how they negotiate human versus AI authorship over time, identifying patterns of protective resistance, pragmatic task delegation, and reflective agency repartitioning. The authors propose 'longitudinal agency partitioning' and suggest design implications for boundary controls, provenance, and authorship norms.
Why it matters
Five-year longitudinal interview data on how creative practitioners actually redraw the human–AI boundary over time is rare; the design implications for boundary controls and provenance transfer to agentic workflow design.
Score 3.80curatedAgent tools & setup
GitHub Changelog · T12026-09-09original ↗ · G·2019
Block pull requests with exposed secrets from merging
Summary
GitHub now lets repository rulesets block pull requests from merging when the PR introduces exposed secrets, scaling secret-detection protections across repositories.
Why it matters
One configuration change in repository rulesets adds an enforceable merge gate against leaked credentials, with no extra tooling required.
Score 3.80curatedAgentic workflow patterns
arXiv cs.HC · T12026-09-03original ↗ · G·1897
Beyond Instruction-Driven Editing: Source-Grounded Problem Discovery with User-Governed Repair for Scientific Posters
Summary
Introduces PROS, an interactive editor for scientific posters that addresses the 'articulation gap' by letting the system surface source-grounded candidate problems while users retain authority over which become repair goals. Includes PROS-Bench (120 papers, 320 PPTX posters) and reports VLM-rated diagnosis quality of 67.2/100 with 87.6% operator-verified target resolution.
Why it matters
Formalises the distinction between problem discovery and committed action in agentic editors, with a concrete benchmark and quantified evidence that the two can diverge.
Score 4.60curatedAgent tools & setup
arXiv cs.MA · T12026-09-02original ↗ · G·1874
Harness Engineering: Anatomy, Architecture, and Evolution of Coding Agents -- A Source-Code Study of Eleven Systems
Summary
A source-code anatomy of eleven production coding agent harnesses (Claude Code, Codex CLI, Gemini CLI, Aider, OpenHands, etc.), mapping seven canonical subsystems with minimal and maximal implementations, cataloging 29 recurring design patterns, and including a longitudinal sample showing convergence over one quarter. Closes with 18 design recommendations and a minimum-viable-harness scaffold.
Why it matters
First systematic source-code dissection of major coding agents across ~4M LOC, yielding concrete patterns and a 90-line scaffold readers can actually build from.

August 2026

Score 4.00curatedAgentic workflow patterns
arXiv cs.MA · T12026-08-24original ↗ · G·1647
Edge-Based Agentic Retrieval-Augmented Generation for Autonomous FHWA Bridge Inspection Compliance
Summary
BridgeGuard is an air-gapped, edge-deployed agentic RAG system for FHWA bridge inspection compliance. It combines vector search over regulatory guides with SQL queries on NBI data, orchestrated by a multi-step ReAct agent. A section-aware chunking method preserves regulatory item boundaries, yielding 94.2% chunk integrity versus 28.4% for naive splitting, and 99.77% accuracy on Delaware's 874-bridge inventory.
Why it matters
Demonstrates a fully offline, multi-step agentic RAG pipeline that meets regulatory compliance at edge scale. The chunking comparison and ablations are directly applicable to other field-based compliance workflows.
Score 3.78curatedAgentic workflow patterns
Addy Osmani · T1.52026-08-14original ↗ · G·1640
Practical Loop Engineering
Summary
Addy Osmani shares principles for engineering AI agent loops, emphasizing well-defined goals and the discipline of retaining human judgment rather than fully delegating decisions to agents.
Why it matters
A senior practitioner's framework for keeping human oversight intact inside agent loops — concrete design pattern, not abstract theory.
Score 4.14curatedAgentic workflow patterns
LangChain Blog · T1.52026-08-11 (crawled)original ↗ · G·1399
How many of your agent's calls actually need a frontier model?
Summary
LangChain benchmarked NVIDIA NeMo Switchyard, an open-source model router, on 145 agent tasks. Only 7% of turns required a frontier model, with a 30B-parameter model handling the remainder. Routing between Nemotron 3.5 Lightning and Claude Opus 4.8 cut total cost by 74% while retaining 93% of Opus's accuracy.
Why it matters
Firsthand benchmark with concrete numbers. Gives teams a defensible cost-optimization strategy for agent LLM calls and a decision formula for adopting model routing.
Score 4.00curatedAgentic workflow patterns
OpenAI News · T12026-08-10original ↗ · G·1375
What building an AI-native finance function taught me
Summary
OpenAI CFO Sarah Friar shares five lessons for building an AI-native finance function, covering automated forecasting, stronger controls, and measuring AI ROI.
Why it matters
Firsthand account from a major tech CFO on operationalizing AI inside a core business function, with concrete patterns other teams can adapt.
Score 3.80curatedPractical tips
harper.blog (Harper Reed) · T12026-08-09original ↗ · G·2313
How to find magical travel experiences using an AI travel buddy
Summary
Harper Reed describes using ChatGPT as a travel planning companion. He shares a taste-profile interview prompt, explains how to feed it into a ChatGPT project, and publishes an installable agent skill (harperreed/travel-agent) for surfacing local spots, with firsthand use in rural Japan.
Why it matters
Firsthand workflow with a copyable prompt and an installable skill file, tested across multiple trips rather than recycled generic advice.
Score 3.80curatedPapers & ergonomics
arXiv cs.HC · T12026-08-07original ↗ · G·1310
Mixed Uncertainty in One View: Co-Visualizing Statistical Variability and Qualitative Confidence
Summary
Three preregistered human-subjects studies (n=923) examine how to visualize qualitative confidence alongside statistical confidence intervals in line-chart forecasts. The work compares text, icons, and integrated encodings such as color, transparency, and a blurred-stroke design, finding that non-textual methods can shape non-expert decision making and yielding design guidelines.
Why it matters
Preregistered design study with a large sample and replicable findings on which visual encodings for qualitative confidence actually shift non-expert decisions, plus concrete guidelines for forecast chart designers.

July 2026

Score 3.80curatedPapers & ergonomics
Human Factors (SAGE / HFES) · T12026-07-27 (crawled)original ↗ · G·1074
A Method for Constructing a Dynamic Seat Adjustment Model to Mitigate Prolonged-Sitting Discomfort
Summary
Peer-reviewed paper proposing a behavior- and height-based dynamic seat adjustment model that generates coordinated low-frequency seat trajectories aimed at reducing subjective discomfort during prolonged public-transport sitting. The model is preliminarily validated.
Why it matters
Primary-source ergonomics research on dynamic seat motion. Useful as a methodological reference for desk-seat or workstation seat designers, though the study context is public transport rather than office work.
Score 3.80curatedPapers & ergonomics
Human Factors (SAGE / HFES) · T12026-07-22 (crawled)original ↗ · G·0942
A Comprehensive Evaluation of Job Rotation: Biomechanical Risk, Body Discomfort, and Psychosocial Demands
Summary
Peer-reviewed Human Factors study comparing a large set of specific job rotation schemes, measuring biomechanical risk, body discomfort, and psychosocial demands to assess effectiveness at reducing musculoskeletal disorder risk among workers.
Why it matters
Empirically compares many rotation schemes rather than endorsing rotation as a blanket control, offering evidence for designing rotation cycles in desk-based and mixed-task work.
Score 4.20curatedAgentic workflow patterns
arXiv cs.MA · T12026-07-21original ↗ · G·0673
FOCAL: Filtered On-device Continuous Activity Logging for Efficient Personal Desktop Summarization
Summary
FOCAL is a privacy-first, on-device multi-agent system that turns continuous desktop interaction streams into task-organized personal logs. A Filter, Brain, Record, and Memory agent cascade cuts token use by 60.4% and VLM calls by 72.3%, and on DesktopBench it holds Task Accuracy at 0.81 under A→B→A interruptions versus 0.03 for the baseline.
Why it matters
Specific multi-agent architecture for on-device desktop summarization, with a benchmark showing task-isolated memory keeps accuracy where a baseline collapses on interruptions.
Score 4.80curatedAgentic workflow patterns
arXiv cs.MA · T12026-07-17original ↗ · G·0516
Does Multi-Agent Debate Improve AI Feedback on Research Papers?
Summary
Pre-registered, identity-masked experiment with 44 economics meta-analysis authors: a single-pass frontier model report was preferred over two multi-agent debate tools, despite debate using roughly thirty times the tokens. AI judges nearly always placed real journal referee reports last.
Why it matters
Empirical counter-evidence to multi-agent debate hype: the simpler single-pass approach won on author-judged usefulness while using 1/30th the compute, and AI judges disagreed with the authors' ranking.
Score 4.00curatedAgentic workflow patterns
martinfowler.com (Exploring Gen AI series, incl. Böckeler) · T12026-07-16original ↗ · G·0692
The Archaeologist’s Copilot
Summary
Article describes modernizing a 20-year-old Java 1.5 codebase using AI assistance combined with stable Docker environments, test-guided incremental refactoring, and evidence-grounded prompts. Early LLM attempts produced plausible but incorrect results; success came from constraining AI with codebase evidence and a step-by-step strategy.
Why it matters
Concrete anti-patterns and a disciplined workflow for using LLMs on legacy code, drawn from a real modernization project rather than theory.
Score 3.78curatedAgentic workflow patterns
Hugging Face Blog · T1.52026-07-15original ↗ · G·0600
Model Routing Is Simple. Until It Isn’t.
Summary
IBM Research describes challenges in model routing for agentic systems. Testing 417 tasks on AppWorld Test Challenge with the same CodeAct agent, Claude Sonnet 4.6 cost $79 total while GPT-4.1 cost $155, despite lower token pricing. The difference stems from cache-hit economics: Sonnet's lower cache-read pricing made it cheaper for context-heavy agent workloads.
Why it matters
Counterintuitive cost data from a 417-task benchmark shows caching, not sticker price, drives real agent-system economics. Reframes routing as systems optimization, not classification.
Score 4.00curatedAgentic workflow patterns
arXiv cs.HC · T12026-07-14original ↗ · G·0369
"Code Is Cheap. Show Me the Talk.": Lessons from Teaching and Managing AI Coding Tool Usage in a Visualization Course
Summary
Retrospective on teaching an upper-level visualization course with AI coding tools, using prompt injections, oral checkouts, and two AI coding labs. Half of students had prior AI use; refinement dominated prompting logs, explanation was almost absent, and 56.4% preferred scaffolded instructions. Final projects were more polished but visually homogeneous.
Why it matters
Concrete management techniques (prompt injections, oral checkouts, scaffolded prompts) backed by real course data — reusable patterns for structuring AI-augmented work or teaching.
Score 3.78curatedAgentic workflow patterns
LangChain Blog · T1.52026-07-13 (crawled)original ↗ · G·0211
Improving Agents is a Data Mining Problem
Summary
LangChain describes treating agent improvement as a data mining problem: mining execution traces to find failures, fine-tuning smaller open models as judges to process trace data at scale, and using evals to hill-climb agent performance through continual learning loops.
Why it matters
Frames agent improvement as an iterative data pipeline rather than prompt tweaking, with concrete hooks for teams already running agents in production.
Score 4.00curatedAgent tools & setup
Anthropic Engineering · T12026-07-13 (crawled)original ↗ · G·0238
Raising the bar on SWE-bench Verified with Claude 3.5 Sonnet
Summary
Anthropic details the agent scaffolding built around an upgraded Claude 3.5 Sonnet for the SWE-bench Verified benchmark, where it scored 49% versus the prior 45% state of the art. The post explains how the agent navigates repositories, edits code, and runs tests to resolve real GitHub issues.
Why it matters
First-party breakdown of how Anthropic structures a coding agent around Claude, including tooling and evaluation harness details that practitioners can adapt.
Score 4.20curatedAgent tools & setup
Anthropic Engineering · T12026-07-13 (crawled)original ↗ · G·0202
Introducing Contextual Retrieval
Summary
Anthropic introduces Contextual Retrieval, a method that prepends chunk-specific context before embedding and BM25 indexing. It reports a 49% reduction in failed retrievals, rising to 67% when combined with reranking, improving RAG system performance.
Why it matters
Primary-source technique with quantified benchmarks and implementation guidance, directly applicable to RAG pipelines.
Score 4.20curatedAgentic workflow patterns
Anthropic Engineering · T12026-07-13 (crawled)original ↗ · G·0199
Effective context engineering for AI agents
Summary
Anthropic's engineering team introduces 'context engineering' as an evolution beyond prompt engineering, framing context as a finite resource that must be holistically curated to reliably elicit desired behavior from LLM-based agents.
Why it matters
Official Anthropic framing of the shift from prompt engineering to context engineering — a paradigm worth tracking for anyone building agent systems.
Score 4.60curatedAgentic workflow patterns
Anthropic Engineering · T12026-07-13 (crawled)original ↗ · G·0185
How we contain Claude across products
Summary
Anthropic describes how it contains Claude agents across claude.ai, Claude Code, and Cowork as agent access expands. It reports that users approve roughly 93% of permission prompts, finding human-in-the-loop supervision insufficient alone, and outlines containment strategies for capping blast radius.
Why it matters
Firsthand engineering write-up with internal telemetry and concrete containment patterns for safely granting agents broad access.