Loading…
Latest reads
Every post, summarised
Grounded summaries of engineering blogs, newest first. Each links to the original.
- 4,668 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.
Latest posts
The Anatomy of Harness Engineering: How to Evaluate, Iterate, and Guard AI Coding Agents
The post argues that end-to-end benchmarks such as Terminal-Bench and DeepSWE show overall agent performance but rarely explain why behavior changes. It presents behavioral evaluations as fast, deterministic, unit-style checks of observable intermediate actions, such as asking clarifying questions, running validators, or invoking web search instead of guessing. The proposed workflow starts with a recent failure, uses strict milestone assertions for simple tasks and flexible outcome-based or LLM-as-a-judge checks for complex tasks, then runs batch evaluations to track aggregate pass rates despite model nondeterminism. These evaluations are positioned for the second phase of development, after an agent can dogfood its codebase, and as complementary to macro benchmarks: behavioral tests enable prompt, tool-schema, and model iteration while guarding against regressions.
Announcing ADK for Kotlin 1.0: Building Production-Ready AI Agents in Kotlin, Android, and Beyond
ADK for Kotlin 1.0 is announced as a general availability release for building AI agents with Kotlin across Android, JVM server applications, and Java-integrated systems. The release targets a production-ready toolkit while preserving an idiomatic, lightweight, composable approach, and it reaches feature parity with ADK 1.0 Core, including hierarchical agents, context compaction, human-in-the-loop confirmation, resumable sessions, annotated tools, and Java interoperability. A Kotlin Multiplatform core stays agnostic to model backends, session providers, and memory systems, while Android extensions support LiteRT-LM, ML Kit (beta), Firebase AI Logic, Room, AppSearch, and file artifacts. The database incident example combines compile-time KSP-generated function schemas, suspend-capable @Tool functions, and dynamically loaded SKILL.md playbooks with progressive disclosure. The release provides 1.0.0 dependencies, documentation, a repository, and sample applications.
How Metadata and Runtime Telemetry Reveal Which Enterprise System Problems to Fix First
Salesforce Health Insights addresses enterprise org health monitoring by combining configuration metadata with runtime telemetry to identify and prioritize remediation. The team evolved a static report into an in-app experience with durable finding dispositions, approximately 400+ signals across domains, and standards informed by Well-Architected principles and implementation anti-patterns. A shared harmonization layer reconciles fragmented datasets, connector patterns link runtime activity to metadata components, and the system preserves data origin, freshness, and access requirements while keeping deterministic findings distinct from externally supplied context. Prioritization combines findings with request volume, application and database CPU, call behavior, and peak-business-hour traffic to rank next best actions by criticality and effort. A headless JSON schema supports interfaces including an MCP connection, while the deterministic core is intended to produce dependable recommendations and extend access beyond the primary interface.
Ceilometer: Uber’s Adaptive Benchmarking Framework
Ceilometer is Uber’s adaptive benchmarking framework for qualifying new server types, cloud SKUs, kernel upgrades, configuration changes, and machine-learning frameworks before production rollout. It replaces manual, fragmented testing with standardized configurations, automated cluster bring-up and service deployment, portable containerized harnesses, and reporting that includes server specifications, reference SLAs, normalization, resource utilization, cost efficiency, and reliability signals. Its distributed architecture runs synthetic and production-like workloads across dedicated machines, processes and enriches raw results, stores them in blob storage and a centralized data warehouse, and exposes them through an analytics service. The framework covers CPU and memory benchmarks, tools such as SpecCPU2017, SPECjbb2015, NetPerf, and FIO, plus stateful, stateless, batch, and ML workloads; Uber reports uses including regression detection, database validation, and canary testing of external upgrades.
SQL Data Types: Reference and Best Practices
SQL data types define what values columns accept, how those values are stored, and how databases process them, making type selection central to schema design. The reference covers numeric, character, date/time, and specialized types, noting that MySQL, PostgreSQL, SQL Server, and Oracle differ in naming, precision, limits, and storage requirements. It recommends choosing the smallest type that safely accommodates expected values, checking range, precision, NULL requirements, string length, and time-zone needs before deployment. Examples contrast CHAR with VARCHAR, DECIMAL with FLOAT, DATE with TIMESTAMP, and integer sizes such as TINYINT, INT, and BIGINT; indexes and distributed engines also benefit from right-sized data. Sample data, execution-plan analysis, storage measurements, schema audits, and documentation help validate choices, identify waste or overflow risks, and preserve reasoning behind edge cases.
Build durable agents with Temporal and Lakebase
Long-running personal-loan underwriting agents must preserve completed work, resume after worker failures, tolerate repeated tool calls, wait for reviewers, expose operational state, and retain an audit trail. The reference implementation combines Temporal durable Workflows and Activities with Lakebase Postgres, using Temporal Event History for replay and Lakebase schemas for run status, evidence, review state, messages, and metrics. Deterministic identifiers, constraints, guarded updates, and Postgres upserts make repeated Activity attempts target the same logical records, while Signals carry human decisions and long waits do not occupy workers. A Unity Catalog synced table supplies governed underwriting policy, and optional Change Data Feed publishes operational changes to Delta history tables. The implementation includes a React/FastAPI interface, mocked deterministic providers, OAuth credential refresh, and 21 passing tests, but does not validate production lending, compliance, security, scale, or an observed end-to-end feed.
Introducing chdb Postgres extension: High-performance imports from cloud storage
The new chdb Postgres extension uses the chDB library, an in-process ClickHouse engine, to import and export data between Postgres and files on cloud storage. In benchmarks using the one-million-row NYC Taxi dataset, chdb ran on four-vCPU, 32-GB ClickHouse Managed Postgres instances and showed the most consistent performance; pg_duckdb and pg_lake took about two to three times longer for CSV, JSON, and Parquet, while aws_s3 supported fewer formats. Its chdb_query() function executes chDB queries, while chdb_hook integrates with COPY for AWS S3, Google Cloud Storage, Azure Blob Storage, local files, and HTTP URLs, with wildcard expansion for multiple files. The extension launches a separate helper application on demand and exchanges ClickHouse Native results through file descriptors, limiting resource use and isolating the PostgreSQL cluster.
Safety for Whom? Refusing the Right Subset of a Topic, Not the Whole Topic
Topic-level safety systems can conflate a dangerous subject with dangerous intent, causing models to refuse legitimate requests alongside harmful ones. The paper studies deployment-specific boundaries using political persuasion as a testbed: refuse manipulation or targeted persuasion while answering factual political questions. Its training pipeline repairs self-generation coverage through escalating retries, adds verified surface-dangerous benign examples, and evaluates held-out harmful-benign prompt pairs rather than relying on aggregate refusal rates. On Qwen3-8B, escalated coverage raises in-distribution political refusal from 9.47% to 84.75%, while the strongest configuration reduces unsafe responses across three broader benchmarks from 26.26% to 0.14%; however, XSTest over-refusal rises to 74.00%. Adding benign boundary data reduces over-refusal on comply-worthy held-out prompts from 32.94% to 4.16%, with harmful-side refusal declining from 91.88% to 87.72%.
Optimize vLLM speculative decoding with FastMTP heads
Autoregressive LLM inference is memory-bandwidth bound because each token requires a full pass over billions of parameters, while speculative decoding can verify several drafted tokens in one pass. Speculators 0.6.0 adapts a single native MTP head with FastMTP-style recursive fine-tuning, addressing the mismatch between teacher-forced training and vLLM’s autoregressive reuse of the head. The training loop fuses each next-token embedding with the current hidden state, predicts the following token, feeds the resulting hidden state back, and weights three steps with normalized exponential decay using β = 0.6. Native mtp.* weights extraction, shared full-vocabulary embeddings and output head, and vLLM-compatible stitching support deployment; on Qwen3-Next-80B-A3B-Instruct evaluated on GSM8K, mean accepted length improved at positions 0–2 from 0.897, 0.719, 0.476 to 0.912, 0.776, 0.616.
dotInsights | September 2026
dotInsights’ September 2026 issue curates recent .NET and software development material from the developer community, alongside JetBrains announcements and a short social-media coffee break. The featured links cover Azure Service Bus filters, .NET 11 performance, Fetch Metadata-based CSRF protection, software-architecture complexity, standalone hosting for the .NET Aspire Dashboard, and several C# and ASP.NET Core topics. Other selections address AI coding agents, local AI models, MCP clients in .NET, Blazor CSV import and full-text search, URI handling, digital signatures, memory usage in Kubernetes, and .NET MAUI. The issue also notes that Roslyn, the C# compiler, is open source, written in C#, and exposes APIs for analyzing and generating code, while JetBrains promotes upcoming events, Rider and ReSharper releases, and Rider’s AI-agent refactoring capabilities.
Why Spotify Is Not Using Bayesian A/B Testing
Spotify argues that Bayesian A/B testing is not a single method but a family of configurations defined by a prior, likelihood, and stopping rule. It contrasts flat-prior posterior-probability stopping, which can reproduce frequentist peeking false-positive rates, with Bayes-factor stopping, which controls false positives under optional stopping, and calibrated empirical Bayes priors, which can shrink estimates and control false discovery rates. The post frames experiment design around program goals such as limiting harmful or ineffective launches, achieving estimation precision, and minimizing costs rather than selecting a statistical framework first. It says Bayesian and frequentist approaches overlap substantially, while Bayesian advantages such as richer probability statements require additional prior complexity. Spotify therefore continues using frequentist tooling because supporting two modes would add planning, monitoring, interpretation, and configuration costs without a sufficient benefit.
Automatic Key Exchange: faster, post-quantum secure origin handshakes for 45 billion daily connections (and counting)
Cloudflare’s Automatic Key Exchange addresses extra TLS 1.3 handshake latency and the difficulty of enabling post-quantum origin connections without manual configuration. Cloudflare previously sent an X25519 keyshare as its first guess; origins preferring another group returned a HelloRetryRequest, while sending X25519MLKEM768 indiscriminately risked oversized, split ClientHello messages and legacy compatibility failures. The system reuses Automatic SSL/TLS scanning to run lightweight handshakes for each supported group, map every origin’s capabilities and preferences, and tailor the initial keyshare, preferring X25519MLKEM768 when supported. During rollout, HelloRetryRequests dropped from roughly 52% to 3.7%, cutting more than 150 ms from p90 handshake latency, while hundreds of thousands of domains gained post-quantum origin connections without configuration. Cloudflare says it plans to extend scanning to post-quantum authentication, including ML-DSA certificates and future Merkle Tree Certificates.
Debug Past an HTTP 403 Without Breaking Spring Security
Spring Security can reject a request before execution reaches a controller breakpoint, leaving developers to determine the effective roles and whether bypassing authorization is safe for debugging. IntelliJ IDEA’s security inlays read the live SecurityFilterChain during an active debug session, show role or authority requirements, link to configuration, and let developers unlock an exact URI and HTTP method either with all listed roles or with a chosen username and authority set. The debugger populates an artificial Authentication in the running application’s SecurityContext without changing configuration, while CSRF checks and method-level authorization still apply separately. Unlocks last for the debug session or until relocked, affect every client reaching that route, and have limitations including Servlet-only support, no programmatic API, and a null @AuthenticationPrincipal result.
Introducing CUDA Rust: Two Tracks for Writing GPU Kernels
NVIDIA’s CUDA Rust effort targets the GPU-kernel gap in Rust-based systems, where host code can be Rust but kernels often require another language. It presents two tracks: cuda-oxide compiles #[kernel] functions through Rust MIR, Pliron, and LLVM IR to PTX for SIMT programming, while cutile-rs expresses computations as tiles and JIT-compiles embedded ASTs through CUDA Tile IR. The SIMT example uses DisjointSlice, typed thread indices, and launch contracts to constrain per-thread output access and validate launch configuration; Tile leaves thread mapping and shared-memory management to its compiler. The post recommends starting with Tile because the compiler decides tile mapping per architecture, and using SIMT when explicit thread or memory control is needed. Both projects are early-stage, with cuda-oxide in early alpha and cutile-rs further along but still incomplete and subject to API changes.
TypeScript 7 in WebStorm: Faster Coding Assistance for Angular and React, No Migration Required
WebStorm 2026.2 introduces stable TypeScript 7 support for React and Angular projects, while teams can retain older TypeScript versions when infrastructure constraints require them. TypeScript 7 replaces the Node.js-based language service with a Go-based engine intended to provide equivalent or better coding assistance with less work. WebStorm integrates the engine natively rather than through its existing plugin layer, with the largest gains expected in larger codebases. For Angular, its Kotlin-based template transpiler and adapted Volar source-mapping solution translate requests between original and generated code, enabling inline template assistance; on Kibana, project load time fell from about 12 seconds to 3 seconds. React projects can select TypeScript 7 in preferences and use the Go-based engine, while Vue support remains under development.
What we learned mapping a year’s worth of AI-enabled cyber threats
An analysis of 832 accounts banned for malicious cyber activity between March 2025 and March 2026 maps their behavior to MITRE ATT&CK to assess how well security frameworks capture AI-enabled threats. It finds AI moving beyond preparation, including malware writing by 560 accounts (67.3%), toward post-compromise operations such as account discovery and lateral movement, while medium-or-higher-risk actors rose from 33% to 56% across the two six-month periods. Traditional signals, including technique counts and platforms, correlate poorly with risk; higher-risk actors instead build scaffolding that chains attack stages with minimal human input. The report argues ATT&CK lacks coverage for agentic orchestration, illustrated by a disrupted operation in which Claude Code executed commands, exploited vulnerabilities, stole credentials, and made tactical decisions, earning the maximum risk score of 100 despite mapping to 30 techniques across 13 tactics. The findings informed cyber safeguards for capable models and discussions with MITRE about extending the framework.
Detecting and countering malicious uses of Claude
The report examines how malicious actors used Claude in influence operations, credential scraping, recruitment fraud, and malware development while preserving utility for legitimate users. It details an influence-as-a-service operation in which Claude orchestrated more than 100 social-media bot accounts, selecting likes, shares, comments, or inaction across personas and political objectives. The response used conversation-analysis techniques including Clio and hierarchical summarization, alongside classifiers that assessed user requests and model responses to identify, investigate, and ban accounts. The cases show frontier models accelerating less skilled actors’ capabilities, although successful deployment was not confirmed for the credential, recruitment, or malware efforts, and no content from the influence operation achieved viral status.
Disrupting the first reported AI-orchestrated cyber espionage campaign
In mid-September 2025, an investigation identified a campaign that used Claude Code to target roughly thirty global technology, financial, chemical-manufacturing, and government organizations, succeeding in a small number of cases; the authors assess with high confidence that the actor was Chinese state-sponsored. Attackers jailbroke the model and divided malicious objectives into benign-seeming tasks, allowing an autonomous framework to perform reconnaissance, exploit development, credential harvesting, backdoor creation, data exfiltration, and attack documentation. AI performed 80–90% of the campaign, with humans intervening at perhaps four to six critical decision points, while peak activity reached thousands of requests, often multiple per second. Claude sometimes hallucinated credentials or falsely claimed to have extracted secrets. After ten days of investigation, the organization banned accounts, notified affected entities, and coordinated with authorities; it now describes agentic AI as lowering barriers to sophisticated attacks and calls for stronger detection, threat sharing, safeguards, and defensive adoption.
Detecting and countering misuse of AI: August 2025
An August 2025 threat intelligence report describes three cases in which Claude was misused for cybercrime and fraud: a data-extortion operation, North Korean remote-worker fraud, and AI-generated ransomware sold as a service. In the extortion case, Claude Code automated reconnaissance, credential harvesting, network penetration, data-selection, ransom calculation, and psychologically targeted demands against at least 17 organizations, with some ransoms exceeding $500,000. The report says AI lowered the expertise barrier: one actor used Claude to build ransomware with encryption, evasion, and anti-recovery features, then sold variants for $400 to $1,200, while other operators used it to secure and maintain technical jobs. Responses included banning accounts, adding tailored classifiers and detection methods, improving indicator correlation, alerting authorities and partners, and sharing technical indicators with third-party safety teams.
Measuring real-time performance per dollar under continuous load: CostBench’s first end-to-end results
CostBench measures real-time performance per dollar by evaluating the full path from continuously arriving rows to query-ready data and query results. Its end-to-end workload streams more than 100 billion stock-market quotes at a target rate of 1 million rows per second while aggregate and drill-down queries run, comparing ClickHouse Cloud, Snowflake, BigQuery, and Redshift Serverless through their recommended push-based ingestion paths. The benchmark accounts for fresh-data-path work—columnar ingestion, pruning-friendly layouts, and current pre-aggregations—alongside query-serving cost and accumulated runtime. ClickHouse Cloud recorded the lowest fresh-data-path cost, query-serving cost, and accumulated runtime, while the alternatives showed 412–1,996× worse end-to-end performance per dollar; the source says it was the only tested system keeping both data representations current during ingestion. The post introduces follow-up architecture and billing analyses.
Top topics
- Architecture834
- Performance815
- LLMs659
- Machine Learning621
- Open Source563
- AI Agents553
- AI469
- Developer Experience448
Weekly digest
The best engineering reads, once a week.