Loading…
Browse by type
Explainer reads
Explainer posts on Yomu, summarised newest first.
- 1,318 posts
- 20 companies
The Good Stuff
Paid picks we use ourselves. Links carry a referral code — a subscription through one earns us a commission at no extra cost to you.
Explainer posts
Building a RAG Pipeline for Semantic Code Search: A Developer Diary and Field Notes
JetBrains Context is presented as a production RAG pipeline for semantic code search, designed to give coding agents precise, citable evidence from repositories instead of relying on keyword tools such as grep. This first installment covers parsing, structure-aware chunking, and vectorization, with the goal of retrieving meaningful functions, symbols, or snippets rather than whole files or insignificant lines. Language parsers produce syntax nodes, and the algorithm retains the largest units that fit, splits oversized leaf nodes when necessary, groups small adjacent units, and keeps documentation, annotations, modifiers, and closing syntax with their constructs. JetBrains Context supports this approach for Kotlin, Java, Python, JavaScript, TypeScript, C#, PHP, Go, and Rust; other languages use line-based splitting. Vectors are generated with an open-weight model on company-operated GPUs, while indexed records store coordinates rather than source code, and the post reports that this setup outperformed hosted embedding APIs on internal code-retrieval benchmarks.
How we AI-updated 1,138 Error-code Docs Pages
Twilio used AI to update 1,138 error-code documentation pages over about 12 weeks, with a human reviewing every change. The team first scored roughly 2,692 pages with a deterministic script covering structure, content, and resources, then routed low-scoring pages to an LLM pipeline while sending borderline pages to human review. That pipeline used OpenAI's gpt-5.4 via the Responses API, generated full-file responses, sanitized them, converted them into git apply patches, linted results, and rejected any page whose code snippet fingerprint changed. Small-batch pull requests, tracking files, local-modification checks, and named reviewers limited duplicate work and kept approval in the workflow. Across about 27 pull requests from March to May 2026, the process marked 1,138 error codes complete, while the team acknowledged that human review was the quality bar and that future scaling would benefit from an automated evaluation harness and diff-based generation.
A multi agent approach audience intelligence
Advertisers often lose strategic intent when campaign briefs become data queries, while planners and data teams also miss insights hidden across large customer datasets. An AI-powered audience generation solution on the Databricks Data + AI Platform addresses this gap by combining Genie, custom tool-calling agents, Agent Bricks, Unity Catalog, and a Databricks App. Planners describe audiences in natural language; Genie produces governed SQL segments, an Affinity Agent calculates statistically validated lift against the full population, and a Supervisor Agent routes or chains these capabilities. The resulting workflow saves audiences in Unity Catalog for auditability and downstream activation, and is described as compressing planning and iteration cycles from days or weeks to minutes while enabling faster response to market shifts and pacing issues.
Worker Backpressure (Part 1)
Canva’s queue workers were designed to consume messages greedily, but that behavior could amplify dependency failures, waste scarce capacity, fill dead-letter queues, and create on-call work. Worker Backpressure adds a local feedback loop to the shared queue library: each processed message reports success or failure, a pluggable controller maintains a 0.0-to-1.0 backoff factor against a configured failure-rate set point, and permits scale concurrency at each poll. The mechanism requires no external coordinator or network calls and adds two arithmetic operations to the runtime path, while trading some throughput for failure containment. In two production incidents, it reduced worker pressure during errors and recovered automatically; one sustained-overload incident recorded a 2.76% average failure rate, 22 DLQ messages over 32.5 hours, and no paging, while rollout to all queue workers remained the immediate goal.
When scanners miss the attack: how Cloudflare Client-Side Security protects storefronts
Modern storefronts can appear healthy while malicious JavaScript diverts affiliate credit, hijacks searches or clicks, tampers with analytics, or loads further code. Cloudflare describes four operations spanning eight payloads that Page Shield ML detected in live traffic; humans verified the findings afterward, while seven payloads were absent from VirusTotal and none received a malicious verdict from URLScan. The detection system represents JavaScript as a graph with a graph neural network (GNN), sends fewer than 0.3% of analyzed scripts to a lightweight LLM on Workers AI, and uses a teacher cohort of frontier models for difficult investigations. This combination handles minification, renaming, some obfuscation, and behavior gated by device, time, geography, referrer, browser state, or cooldowns. The post concludes that continuous browser visibility and runtime context are needed because static, one-time scans can miss selectively executed attacks.
How Data 360 Builds Trusted Context: The Enduring Layer for Enterprise AI
Enterprise AI needs a way to assemble the smallest sufficient set of current, relevant, and authorized evidence for each agent turn instead of sending an entire data estate to a model. Salesforce presents Data 360 as a shared Trusted Context runtime spanning Salesforce, external platforms such as Databricks and Snowflake, operational systems, documents, knowledge bases, and live signals. Its progressive pipeline connects data through ingestion and Zero Copy, then harmonizes, unifies, models, retrieves, reconciles, governs, and compiles evidence into a token-fit Context Pack with citations and action contracts. The Agent Context Engine also returns typed interaction traces, promoting only approved facts into durable memory and validated outcomes into learnings. Q2 FY27 production usage is reported at 104.3 trillion connected records, including 81.5 trillion Zero Copy rows, while proposed benchmarking covers correctness, authorization, provenance, freshness, latency, cost, and memory continuity.
"Regex for Rows": Simplifying Pattern Detection in SQL with MATCH_RECOGNIZE
SQL's unordered-set model makes it difficult to detect ordered event sequences such as consecutive login failures followed by a success, especially when timing matters. MATCH_RECOGNIZE, now available in Databricks compute including Lakehouse Real-Time, expresses row sequences through regex-like states while supporting time anchoring and row-relative checks. The post shows its use for credential-stuffing anomalies, V-shaped stock reversals, checkout abandonment, and equipment-failure signals: FIRST can anchor a failure window, the end-of-partition marker $ can represent inactivity, and PREV/NEXT can compare sensor readings. By replacing chains of common table expressions, gaps-and-islands scaffolding, self-joins, and window-function logic, the clause is presented as easier to validate, maintain, and update for event-sequence analytics.
Behind the Scenes: How the OpenTelemetry Plugin Maps Your Microservices in Real-Time
Outdated architecture diagrams and static code analysis can miss how services are connected at runtime, so the OpenTelemetry plugin builds a Service Map from live telemetry. With a lightweight local OpenTelemetry backend, the plugin supplies standard OTel environment variables, receives application spans, and continuously updates an internal architecture model that the IDE renders on demand. Its stream-processing algorithm classifies spans through semantic attributes, identifies services, matches HTTP CLIENT and SERVER spans across propagated trace context, and infers database or messaging dependencies while retaining unmatched spans in memory. Because spans can arrive out of order and traces have no definitive finish signal, the map is incrementally reconstructed rather than built from complete traces; the result is a language- and vendor-agnostic view for OpenTelemetry-instrumented applications across several JetBrains IDEs.
I got frustrated with AWS, so we built our own sandbox
At a healthcare startup, an agent workflow built with LangGraph was constrained by context length and by the limited tools exposed to it for data analysis. Changing one node gave the agent access to Amazon Bedrock AgentCore Code Interpreter, allowing it to use pandas and other available data-science libraries in a sandbox. It could inspect files and outputs, write and run code, and adjust its analysis iteratively while keeping working data and intermediate artifacts in the workspace rather than sending every row through the conversation. Although the experiment demonstrated the usefulness of this controlled autonomy, regional availability forced the author to use a US region from Malaysia, motivating BoxCompute, an infrastructure project intended to provide agent workspaces with control over execution, resources, output collection, and cleanup.
How Value Induction Reshapes LLM Behaviour
Conversational LLMs are post-trained on language expressing traits including curiosity, open-mindedness, and empathy, alongside values such as helpfulness, harmlessness, and honesty. Because values are complex and inter-related, inducing one may alter others and may also affect safety, anthropomorphic language, addictiveness, or sycophancy. The authors fine-tune models with curated value subsets from existing preference datasets, then measure changes in other value expressions, safety, anthropomorphic language, and performance on various question-answering benchmarks. They report that induction can express related and sometimes contrastive values, while positive-value induction increases safety. Across the values studied, it also increases anthropomorphic language, making models more validating and sycophantic.
Glyph: A Multi-Strategy Agentic System for Column Description and Sensitivity-Ontology Tagging of Enterprise Data Catalogs
Glyph addresses enterprise data lakes where tables accumulate faster than human stewards can document or classify them, leaving column descriptions and governance labels incomplete. It models column description generation and sensitivity-ontology tagging as cooperating LLM agents orchestrated as stateful graphs. The Descriptor uses active Retrieval-Augmented Generation to retrieve pipeline source code from an enterprise GitHub, while the Tagger runs description, line-of-business regex, and metadata strategies in parallel before combining ranked outputs with Reciprocal Rank Fusion. A fine-tuned six-layer MiniLM metadata encoder raised same-tag retrieval on an in-distribution held-out split from NDCG@10 0.55 to 0.92 and MAP@100 0.19 to 0.90, relative to the stock base encoder. Glyph also evaluates end-to-end multilabel tagging with a recall-weighted F2 objective across three groups and an ablation, emphasizing value-free, code-grounded operation, per-tag provenance, and graceful degradation for auditable production cataloging.
Large-Scale Automated Dependency Analysis Across Uber's Service Mesh
Uber describes an automated method for cataloging dependency behavior across its service mesh, where nested asynchronous calls make failure attribution difficult. Because distributed traces are sampled and erroneous requests are rare, trace-based correlation can take too long; the system instead instruments yarpc middleware to observe every request and correlate inbound failures with outbound failures at service-and-endpoint granularity. An in-memory request record links each inbound request to its outbound calls, while emitted metrics capture caller and callee nodes and whether each errored. The resulting edge probability, calculated as caller failures divided by requests where the callee failed, classifies dependencies as fail-close at Pc >= 0.8, fail-open at Pc <= 0.2, or unknown between those thresholds. Uber says the initiative cataloged dependencies of critical APIs and indexed most fail-close anomalies, supporting reliability work and regression guardrails for Failover Architecture 2.0.
Beyond the Model: Engineering AI Infra with Scientific Judgement
Airbnb describes Insight Miner, an agent harness designed to make unstructured data investigations scalable, reproducible, and audit-ready. Its problem was a bespoke process spanning data discovery, expert review, representative datasets, and manually curated methods that could not keep pace with new languages, geographies, and LLM-based products. The shared package combines extraction, embedding, and clustering with prompt tuning, hard-example mining, contrastive labeling, and a path from unsupervised exploration to classification, while an agent handles research questions, execution, and methodological guidance. Airbnb says investigations that once took months can now take days, allowing mechanical work to scale while data scientists spend more time inspecting ambiguous evidence, testing hypotheses, and improving techniques. The tool also lets nontechnical experts run analyses across use cases including survey coding, model evaluation, and fraud patterns.
Dense vs. MoE Models: Active Parameters, Throughput, and When to Choose Each
Dense models activate every parameter for every token, while Mixture-of-Experts (MoE) models route each token through a selected subset of feed-forward experts. This changes the relationship between capacity, memory, and compute: total parameters determine the memory required to load a model, whereas active parameters largely determine per-token work. At equal total size, MoE models often provide higher throughput, especially at batch size one, but routing, simultaneous expert residency, and reduced KV-cache room add serving complexity; latency advantages narrow at high concurrency. The comparison uses Nemotron 3.5 Lightning, a roughly 30B-parameter Mamba-2 plus MoE plus Attention hybrid with 3B active parameters, alongside dense and MoE examples, and concludes that dense models suit simpler serving and fine-tuning while MoE models suit throughput-oriented deployments with manageable memory and operational complexity.
How NVIDIA Groq 3 LPX Deterministic Execution Drives Power-Efficient High-Interactivity Inference on NVIDIA Vera Rubin
AI factories are constrained by site power, making performance per watt especially important for high-interactivity inference. The NVIDIA Groq 3 LPX rack uses deterministic execution: its compiler schedules computation, data movement, and clock-cycle timing across 256 LPU chips, then derives a cycle-level current-demand curve. Preemptive Power (PEP) prepares the power-delivery network for upcoming spikes, while Clock Period Synthesis (CPS) lengthens or shortens selected cycles to reduce di/dt, voltage droop, and the voltage guardband. Internal testing reported more than 60% less voltage drop, while the source estimates a high-single-digit baseline-voltage reduction and potentially low-double-digit workload-power reduction; paired with Vera Rubin NVL72, the platform is described as delivering up to 35x higher throughput per megawatt than GB200 NVL72 for 2T+ parameter models at long context and high interactivity.
How NVIDIA NVLink 6 Delivers Multi-Layer Resiliency for AI Factories
NVIDIA NVLink 6 is presented as a scale-up networking fabric for large AI factories, where dropped packets, link degradation, and node interruptions can reduce training goodput or inference capacity. It connects 72 Rubin GPUs in the Vera Rubin NVL72 into a single scale-up domain and uses a multi-layer resiliency stack spanning physical, link, application, software, and system layers. Physical Layer techniques include lightweight FEC, Physical Layer Retry (PLR), UPHY recovery, and hardware replay buffers, while credit-based flow control (CBFC) prevents injection when the next hop lacks buffer capacity. Redundant switch trays and management paths, local fault containment, NMX high availability, checkpointing, topology-aware recovery, and targeted switch maintenance are described as mechanisms for containing failures and preserving workloads; the post concludes that this integrated design targets continuous operation and maximum cluster productivity.
Your Agent Aced the Task. Will It Do It Again?
Agent reliability can be obscured by average benchmark scores: Mean@k reports average success, while Pass^k measures tasks succeeding on every repeated run. On AppWorld, a ReAct agent using GPT-4.1 achieved Mean@5 of 77.4% but Pass^5 of 53.0%, leaving a 24.4-point consistency gap, with temperature set to 0.0. The Consistency Analyzer diagnoses flip-prone decision points by resampling each step in one recorded trajectory with k=5 completions, requiring no ground truth, logits, or end-to-end replay. ALTK-Evolve turns flagged instability into reusable consistency guidelines that are retrieved at inference time. In the reported evaluation, these guidelines raised aggregate Pass^5 to 69.0% from 53.0% and Mean@5 to 81.0% from 77.4%, cutting the gap to 12.0 percentage points; similar-task Pass^5 also improved by 13.0pp.
Scaling Federated Learning Across Docker, Kubernetes, and Slurm with NVIDIA FLARE
NVIDIA FLARE addresses the operational challenge of running federated learning across sites that use different infrastructure, including Docker, Kubernetes, and Slurm. Its two-layer architecture keeps long-running server and client parent processes available for authentication and coordination while launching transient job workers through each site’s native runtime. Jobs declare portable GPU, CPU, and host-memory requirements, which local launchers translate into Docker containers, Kubernetes pods, or Slurm allocations; studies add scoped membership and site-owned mappings for datasets, images, secrets, and scheduling policy. The result is a heterogeneous federation where organizations retain local resource authority, can allocate compute on demand, and run multiple studies without requiring one shared platform.
Java 27 in IntelliJ IDEA
Java 27, released on September 15, 2026, follows Java’s six-month release cadence and includes nine JEPs: four final features, four previews, and one incubator feature. IntelliJ IDEA supports Java 27 from day one, with setup through Project Structure, automatic JDK configuration from .sdkmanrc or .tool-versions, and language-level controls for stable and preview features. Stable changes include G1 as the default garbage collector in all environments, hybrid post-quantum key exchange for TLS 1.3, compact object headers by default, and JFR in-process data redaction. Preview and incubator work covers lazy constants, primitive types in patterns, instanceof, and switch, PEM cryptographic encodings, and the Vector API, while faster debugger stepping is available through a JDK 27 fix also backported to JBR 25.
Red Hat Developer Hub: Preventing compliance violations with AI coding agents
AI coding agents can generate services quickly, but they may miss organization-specific architecture, ownership, and compliance rules stored in an internal developer portal. A demo connects Claude Code to a Backstage-based Red Hat Developer Hub through Backstage CLI actions, allowing the agent to query domains, components, APIs, software templates, and TechDocs before proposing a design. In a simulated Parasol Insurance catalog with more than 500 entities across 14 domains, the agent finds an existing but unrelated reconciliation service and resolves a Claims–Billing messaging conflict by keeping IBM MQ at the payment boundary and publishing internal events to Kafka. It then uses catalog relationships and specifications to identify integrations and produces a Technology Decision Summary that cites the sources behind each choice, making the resulting decisions reviewable and traceable.
Top topics
- Architecture830
- Performance812
- LLMs648
- Machine Learning617
- Open Source561
- AI Agents542
- AI454
- Developer Experience446
Weekly digest
The best engineering reads, once a week.