Loading…
Latest reads
Every post, summarised
Grounded summaries of engineering blogs, newest first. Each links to the original.
- 4,718 posts
- 20 companies
Loading…
Latest reads
Grounded summaries of engineering blogs, newest first. Each links to the original.
The Supabase .NET SDK provides a single C# client for Auth, Database, Storage, Realtime, and Edge Functions; version 8.0.0 moves serialization to System.Text.Json and adds injectable HttpClient with retry. It also ships a dependency injection package for ASP.NET Core that registers the client as scoped via IHttpClientFactory to pool connections. Models derive from BaseModel with table and column attributes; queries use From () with Where filters, and updates require explicit calls without a change tracker, supporting both model-based and query-builder approaches. Authentication covers email/password, OAuth, magic links, phone, and SSO, with auto-refresh tokens attaching to subsequent requests for Row-Level Security. Setup for ASP.NET Core uses AddSupabase, while other apps construct the client manually and call InitializeAsync; the post also distinguishes publishable and anon keys for client apps from secret and service keys that bypass Row-Level Security.
The Supabase .NET SDK v8 is now available, consolidating all packages under a single repository with a unified version number. Previously, packages like Gotrue, Realtime, and the main package were scattered across multiple repositories with separate versions and lifecycles, making upgrades complex. From v8.0.0, all packages—Gotrue, PostgREST, Realtime, Storage, Functions, and the umbrella package—share the same version, and previous repositories are archived as read-only. The jump from v1.x.x to v8.0.0 was necessary because Realtime was already on v7.x.x, and this ensures lockstep versioning going forward. A new dependency injection extension package registers the Supabase client with the .NET service container and lets IHttpClientFactory manage its HttpClient, starting at v8.0.0 as well. The update switches serialization from Newtonsoft.Json to System.Text.Json, adds configurable retry with backoff, accepts publishable and secret API keys, and targets netstandard2.1, with a migration guide available for breaking changes.
Databricks announced that Spatial SQL and spatial joins now run on Lakehouse//RT, delivering up to 38X faster performance under high concurrency compared to traditional cloud data warehouses. The platform supports GEOMETRY and GEOGRAPHY data types, over 100 spatial functions, and high-performance spatial joins, all within the open lakehouse using Apache Spark and Parquet files. Organizations in insurance, agriculture, energy, telecom, retail, and logistics can consolidate geospatial workloads onto a single governed platform, eliminating separate stacks like PostGIS. Benchmarks using SpatialBench show Lakehouse//RT is 3X faster on average and up to 20X faster for single-query concurrency, with no code changes required. The service is available in Beta, allowing users to run existing spatial queries directly on Delta and Iceberg tables without copying data.
Atlassian faced a security data pipeline ingesting 44 billion events daily from 20 years of growth and acquisitions, each with proprietary logging formats. To unify this fragmented telemetry, the team standardized on the Open Cybersecurity Schema Framework (OCSF) and built a security lakehouse on Databricks. They embedded the entire OCSF schema into Go structs (TypeGo), guaranteeing compile-time correctness and reducing AI hallucinations during automated transformation. Table optimization with partitioning and Z-ORDER cut query times by 95%, and Graviton clusters with Photon reduced cost and runtime by two-thirds. The result was a self-service platform where SOC analysts can onboard new data sources in days, and detection teams moved from one-to-one alerts to signal-based clustering using standardized OCSF values.
Databricks’ AI Research team uses AI Runtime (AIR) to manage hundreds of parallel research workstreams for products like Genie, Smart Routing, and AI Functions, training models from a few billion to nearly a trillion parameters. AIR performs preflight checks on GPU nodes to ensure health and communication before training, and monitors node health during execution to detect issues like NCCL failures or thermal throttling. The AIR CLI and custom Docker containers enable fast experimentation by launching jobs in seconds without rebuilding environments, supporting both supervised fine-tuning and reinforcement learning. MLflow integration automatically tracks experiments, while Unity Catalog FUSE mounting gives governed access to research data across workspaces without changing training code. The team has scaled to agent-led training, where AI agents propose changes, launch trials, and evaluate results, increasing training job volume by 8x across AI Functions research workloads.
Anthropic is offering Claude for Enterprise and Claude for Government to all three branches of the U.S. government for $1, removing cost barriers to AI adoption. The offer builds on Claude’s addition to the GSA schedule and supports FedRAMP High workloads for sensitive unclassified data. Government partners already using Claude include the Department of Defense, Lawrence Livermore National Laboratory with 10,000 daily users, and the District of Columbia Department of Health. The package includes frontier models with continuous updates and technical support for one year, with coordination underway with GSA for implementation. This expanded access aims to accelerate AI adoption across the federal government for scientific research, national security, and constituent services.
Granite Switch 4.1 models combine chat, general and specialized decisions, and speech-to-text from a single hosted endpoint. The 8B and 30B models include a new Decision aLoRA for general-purpose decision-making and a Guardian aLoRA that outperforms Jev 1.13 and other closed models on safety tasks. They also embed Granite-Speech-5.0-Turbo-CTC, the fastest ASR model on the OpenASR Leaderboard, enabling efficient transcription. All components are composed into one model using cache-efficient Activated LoRAs, taking up about the same GPU VRAM as the base model alone, and the KV cache from base generation can be reused for any subsequent adapter decisions. Granite Switch 4.1 can switch between chat, decision, and spoken requests quickly, and a 3B variant achieved 60-100 decisions per second playing Doom, compared to Jev’s 10, while also narrating and answering questions.
Meta built Forum using AI-native engineering, treating an AI agent as a development participant rather than a faster typing tool. The team shifted from human-in-the-loop to human-on-the-loop, where the agent autonomously plans, implements, builds, tests, and corrects its work while the engineer sets goals and approves outcomes. Success was measured by verified, production-ready results, not code volume, with automated verification signals like build status and unit tests filtering output before review. Meta adapted the codebase for legibility, gave agents tools to inspect live view hierarchy directly instead of descriptions, and created a separate sandbox build to isolate prototype code from production. This approach helped Forum reach core feature parity with Facebook Groups in eight weeks and caught a memory issue during development that roughly halved memory use and page faults.
The 2026 FIFA World Cup produced the largest planned stress test of Meta’s infrastructure, with WhatsApp setting an all-time record of over 30 million messages per second during the final and delivering 8.4 billion messages in its peak five-minute window. Traffic analysis revealed that surges in sharing primarily occurred during breaks and final whistles, not during active gameplay, shaping capacity decisions. Instagram and Facebook saw live video watch time spike 90% and 242% respectively, while recommendation traffic rose 80% week-over-week. To absorb these spikes, Meta layered pre-provisioned capacity buffers, country-level graceful degradation, cascade prevention, and early warning signals across all serving tiers. Notably, the team did not shift load between regions or freeze deployments, relying instead on year-round reliability enhancements and phased runbook execution.
FactSet’s Text-to-Formula (TTF) system converts natural-language investment queries from professionals into correct FactSet Query Language (FQL) formulas. Initially built on retrieval-augmented generation (RAG), the approach hit a quality ceiling as queries became more specialized and formulas more complex. FactSet adopted fine-tuned open models within a Compound AI architecture, reducing end-to-end latency by about 60 percent and improving quality for complex formulas. Fine-tuning runs on Databricks AI Runtime with serverless GPU infrastructure, where engineers choose compute based on prompt length and output needs, and the pipeline integrates with MLflow, the Lakehouse, and Lakeflow Jobs for a unified workflow. This setup cut model setup, training, and deployment from days to hours, enabling the team to retrain every two to three weeks as new data and models become available.
Meta launches Muse Spark Open, an 816B-parameter mixture-of-experts model with 42B active parameters and a 1M token context window, trained on NVIDIA GB300 NVL72 for agentic coding workflows. It plans, acts, observes, and self-corrects across multistep tasks like multifile refactoring, using tool calls and subagent coordination while holding an entire project in context without breaking it into pieces. Its 1.8B perception encoder natively processes interleaved text, images, video frames, and audio in a single pass. On GB300 NVL72 with speculative decoding, it achieves over 50K tokens per second per GPU and over 600 tokens per second per user. Sparse MoE routing activates only 42B parameters per token, and NVFP4 quantization reduces memory footprint for larger KV cache. The model is available via Hugging Face and NVIDIA NIM, with support from SGLang, vLLM, and TensorRT LLM for serving, and fine-tuning through NeMo AutoModel and Megatron Bridge.
Managing outside collaborator access and stale repositories in a shared GitHub organization becomes harder to keep up with as projects grow. Evergreen is a GitHub App built with native GitHub features—scheduled actions, webhooks, issues, and custom properties—that automates two workflows: quarterly outside collaborator access reviews and stale repository detection and archival. Its design principles are fail secure, revoking access if no one responds, and never destroy anything, keeping archived repos recoverable and every change tracked in issues. The app is a Go webhook server that stores state as JSON files in a controller repo, using git history as the audit trail, and interacts via checklist issues in each repository. Rollout included dry-run mode, overcommunication, Splunk observability, and proactive rate-limit pacing; the pattern is adaptable using native GitHub features.
Anthropic has begun providing employee-like access to METR, a nonprofit evaluator of frontier AI models, as part of its commitment to independent third-party assessment. The company signed the White House Accord on Super Intelligence and aims to close the gap between what evaluators and the company itself understand about its systems. To evaluate frontier models, evaluators require expertise in capabilities assessment, independence, deep access to training and safeguards, and freedom to publish results. Anthropic defined three tiers of access: the first tier allows publication only with company approval, the second tier permits publication without approval but with possible redactions, and the third tier grants employee-like access with similar publication rights as the second. METR is currently at the third tier, while SecureBio operates at the first tier, and collaborations with Accenture and Faculty are in early stages.
Traditional secure web gateways enforce security through a binary allow, block, or log model, which limits customization for complex environments. Cloudflare is integrating its Cloudflare Gateway and Data Loss Prevention with Cloudflare Workers, allowing customers to run arbitrary code inline at the moment a security decision is made. The new Forward action, available in public beta, routes matched HTTP traffic to a Worker that acts as inline middleware, enabling request and response modification or asynchronous tasks like sending Slack messages. Custom Functions, arriving before the end of the year, call a Worker during policy evaluation and return values used as selectors, enabling data-conditional rules and DLP validation against external APIs. Because Gateway and Workers run on the same machines in Cloudflare’s global network, dispatching traffic to a Worker is a local function call with no hairpinning or added latency.
Many engineering teams mistakenly think building an internal AI knowledge context layer is as simple as standing up a vector database and wiring a RAG pipeline to documentation. In reality, a production-grade pipeline requires five continuously maintained stages—capture, deliver, govern, organize, and validate—where each stage’s long-term cost compounds as vendor APIs change and content ages. The critical innovation is the knowledge node, a living record carrying source, ownership, and an automatically computed trust score based on authorship, usage, recency, and conflicts. Trust decays over time, prompting owners for lightweight revalidation when confidence drops, and knowledge is automatically scoped to the correct audience without manual filing. For product companies not in the knowledge-infrastructure business, the compounding maintenance burden makes building this pipeline a distraction compared to leveraging existing platforms like Stack Internal.
Sapienz, a multi-objective automated testing system for Android, received the ISSTA 2026 Most Influential Paper Award for its 2016 research by Ke Mao, Mark Harman, and Yue Jia. The system combines three goals: exploring more of an app, finding more crashes, and reducing the steps needed to trigger them, making failures easier to investigate. In an evaluation of 1,000 popular Google Play apps, Sapienz found 558 previously unknown crashes. Within 17 months of publication, it was deployed at Meta, where it now runs millions of tests daily, surfaces thousands of unique crashes each week, and helps safeguard hundreds of app releases each month. Sapienz also became the app-exploration engine for dynamic taint analysis at WhatsApp using synthetic conversation test accounts.
Dynata, a large first-party data company, sought to accelerate delivery across distributed teams while improving cross-functional alignment. They adopted a staged AI approach: first AI-assisted coding via Amazon Q Developer, then connecting AI to project context from Jira, Confluence, and Lucid, and finally implementing the AI-Driven Development Lifecycle (AI-DLC) methodology using Kiro IDE. AI-DLC, powered by Amazon Bedrock, structures delivery into Inception, Design, and Construction phases with Specs, steering files, and hooks. This process forced all teams to agree on requirements and design during a Day 1 workshop, surfacing gaps before coding began. Consequently, features that once took 3–4 sprints were completed in about one sprint, cutting development phase time by roughly 70% and reducing post-launch defects.
The Kotlin Foundation wrapped up its fourth Google Summer of Code program in 2026, supporting two projects by Ayako Hayasaka and Mark Seif Khalil. Ayako implemented tail-call support in the Kotlin/Wasm backend, using return_call and return_call_ref instructions to avoid stack growth in recursive programs, and then found and fixed a missing optimization for return_call in the V8 engine that caused a performance regression. Mark built a database and automated pipeline to collect and structure scattered information about where Kotlin is taught worldwide, covering courses, curricula, repositories, and learning platforms. In Ayako’s benchmarks, several recursive patterns achieved 1.7 to 4.3 times faster execution at a recursion depth of 10,000, and at a depth of one million, the same programs crashed without tail calls but completed successfully with them. The education landscape project now provides a reusable foundation for monitoring Kotlin’s presence in education and identifying trends and gaps.
Red Hat AI released optimized FP8 and DSpark speculative decoding checkpoints for the Nemotron 3.5 Super VL 120B multimodal mixture-of-experts (MoE) model. These checkpoints are designed to double inference throughput on NVIDIA Hopper GPUs without compromising output accuracy. FP8 quantization uses LLM Compressor to reduce model weights and activations from BF16 to 8-bit floating point, cutting memory and storage requirements in half while preserving quality. DSpark speculative decoding proposes 8-16 tokens in a single forward pass and uses a lightweight Markov head to introduce dependencies between speculated tokens. Benchmarks on HumanEval, math_reasoning, and RAG datasets show that each technique delivers meaningful gains individually and combining them outperforms any intermediate configuration at every concurrency level.
NVIDIA released Nemotron 3.5 Super VL, an open 120B-parameter multimodal model for specialized enterprise agents. Its mixture-of-experts architecture activates 12B parameters per token, balancing capacity and inference efficiency. The model combines reasoning, instruction following, tool use, and understanding of text, images, documents, and video for tasks including coding, web research, document intelligence, and cybersecurity. On a single GPU, it achieves up to 5x higher throughput for agentic tasks than competing open models, and it scores higher than the larger Nemotron 3 Ultra on Hermes Bench and LangChain evaluations. Early adopters report cost reductions of 40x for credit processing and 10x lower latency for receipt parsing, and organizations can post-train the model with NVIDIA NeMo for domain-specific workflows and deploy it with NVIDIA NIM and Dynamo, which serves up to 8x more users with 85% lower first-token latency.
The best engineering reads, once a week.
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.
