Loading…
Latest reads
Every post, summarised
Grounded summaries of engineering blogs, newest first. Each links to the original.
- 4,671 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
How Notion catches breaking schema changes before they reach production
Schema drift in typed API requests and queue tasks can cause incidents when contracts change without detecting whether old payloads remain valid. Notion's CI guardrail extracts schema types from the base branch and compares them with PR types using TypeScript's compiler API, checking whether the old type is assignable to the new one. The check auto-discovers registered API endpoints and queue task types, uses git archive for base files, and currently covers approximately 1,300 endpoints and 296 task types. Backward compatibility blocks changes that reject old data, while queue schemas also receive an advisory forward check for rollout ordering. The rollout began with warnings, then made backward checks blocking and added a GitHub label for intentional exceptions; type checking still cannot catch extra runtime validation or semantic behavior changes.
100,000 GitHub stars
Supabase reports reaching 100,000 GitHub stars and says eight million developers are building with the platform. Founded in 2020 by two people around a Postgres database, the project says it optimized for developers having a good time rather than for star counts, treating the milestone as evidence of a community that contributed, stayed involved, and built unexpected projects. Its open-source approach centers on supporting existing tools and communities: projects run unmodified Postgres, PostgREST handles billions of API requests daily, pgvector stores embeddings, Deno underpins Edge Functions, imgproxy handles image transformations, and Phoenix supports Realtime at almost 1.5 billion messages per day. The team says it aims to help make Postgres the default database for new internet projects, including by pursuing petabyte scale and pluggable storage.
Meet the new Cursor
Cursor 3 is a unified workspace for building software with agents, introduced as Cursor moves toward a third era in which fleets of agents work autonomously to ship improvements. Built from scratch around agents rather than as an extension, its multi-workspace interface brings local and cloud agents into one sidebar across multiple repositories and supports parallel work. Sessions can move from cloud to local for desktop editing and testing, or from local to cloud for offline and longer-running work; cloud agents also produce demos and screenshots for verification. The new diffs view streamlines editing and review through staging, committing, and PR management, while files with full LSPs, an integrated browser, and Marketplace plugins retain IDE capabilities; users can switch back to the Cursor IDE. Cursor says these foundational pieces—model, product, and runtime—are in place for more autonomous agents and team collaboration, while IDE investment continues until codebases are self-driving.
Welcome Gemma 4: Frontier multimodal intelligence on device
Google DeepMind’s Gemma 4 family is available on Hugging Face as Apache 2-licensed, multimodal models intended for use across agents, inference engines, fine-tuning libraries, and on-device deployments. The family accepts image and text inputs across five sizes, while E2B, E4B, and 12B Unified also support audio; context windows range from 128k to 256k tokens. Its design combines alternating local and global attention, dual RoPE configurations, Per-Layer Embeddings, shared KV caches, and model-specific vision and audio handling, including an encoder-free unified 12B variant. The post documents integrations for Transformers, llama.cpp, MLX, WebGPU, Rust, and other tools, alongside fine-tuning examples and demos. Reported instruction-tuned results include a 1452 estimated text-only LMArena score for 31B and benchmark results spanning reasoning, coding, vision, audio, and long-context tasks.
PostgresBench: A Reproducible Benchmark for Postgres Services
PostgresBench is a public, reproducible benchmarking suite created to compare transactional performance across managed PostgreSQL providers. The benchmark uses the standard pgbench tool with a TPC-B-like write-intensive workload executing across 256 clients and 16 threads for ten-minute durations. Evaluators tested five services—including Postgres managed by ClickHouse, AWS Aurora, AWS RDS, Neon, and Crunchy Bridge—under 100 GB and 500 GB scale factors using default configurations in the us-east-2 region. Across tested configurations, Postgres managed by ClickHouse delivered higher transactions per second and lower P99 latencies due to physically co-located NVMe primary storage. These results demonstrate that remote network storage layers, such as EBS volumes, introduce significant latency overhead during sustained write and fsync operations.
Rill and ClickHouse: real-time operational BI for a metered world
Granular operations across cloud computing, payment platforms, and artificial intelligence workloads produce high-volume event streams that traditional business intelligence tools struggle to unify and aggregate. Operational BI platform Rill addresses this challenge by combining declarative data loading through the dlt framework with ClickHouse's high-performance analytical aggregation engine. Data transformations, SQL-based measure expressions, and dashboard configurations are defined entirely as code using SQL and YAML. This declarative, metrics-first architecture allows developers to validate analytics locally, track changes in Git, and eliminate intermediate caching layers when querying live data. By querying ClickHouse directly and linking AI-generated insights back to underlying queries, the system provides traceable, real-time operational analytics for enterprise environments.
Announcing Role Based Access Control in ClickStack
Managed ClickStack has introduced Role-Based Access Control (RBAC) to replace a legacy model where all users shared identical instance-wide permissions or required duplicate HyperDX instances for isolation. Integrated with ClickHouse Cloud for identity management, the new system allows administrators to assign users to roles governed by three access tiers: no access, read, and manage. Permissions apply across platform resources including dashboards, saved searches, sources, alerts, webhooks, and notebooks, with support for fine-grained rules matching names, IDs, or tags. While table-level source restrictions work through ClickStack, row-level access currently depends on separate ClickHouse SQL console roles, which the team intends to unify directly inside ClickStack in future releases.
Falcon Perception
Falcon Perception is a 0.6B-parameter early-fusion Transformer for open-vocabulary grounding and segmentation from natural-language prompts. It processes image patches, text, and task tokens in one sequence, using hybrid attention so image tokens receive bidirectional context while text and task tokens decode causally. Its Chain-of-Perception interface predicts each instance as coordinate, size, and segmentation tokens, with specialized Fourier-feature and dot-product heads for dense outputs. The post also introduces PBench, which profiles attributes, OCR-guided identification, spatial understanding, relations, and crowded scenes. Falcon Perception reaches 68.0 Macro-F1 on SA-Co versus 62.3 for SAM 3, while presence calibration remains weaker at MCC 0.64 versus 0.82; the 0.3B Falcon OCR model scores 80.3 on olmOCR and 88.6 on OmniDocBench.
Supabase docs over SSH
Supabase is launching supabase.sh, a free public SSH server that gives AI coding agents bash access to Supabase documentation as a public experiment. Its virtual file system exposes every documentation page as a Markdown file under /supabase/docs, allowing agents to use familiar commands such as grep, find, ls, cat, head, and tail to browse, search, and skim current content. The service runs on just-bash, Vercel’s open-source Bash emulator written in TypeScript, inside a sandboxed JavaScript runtime over a virtual file system, so it does not execute an actual shell on the server. Project instructions can be added with ssh supabase.sh agents >> AGENTS.md or generated through ssh supabase.sh setup | claude; the post contrasts this exploratory workflow with the existing search_docs MCP tool and presents supabase.sh as an early experiment.
Any Custom Frontend with Gradio's Backend
gradio.Server enables applications built with React, Svelte, or plain HTML/JavaScript to use Gradio’s backend facilities instead of Gradio components. The post demonstrates this with Text Behind Image, whose browser UI layers a background, styled text, and a transparent foreground produced by BiRefNet, then exports a PNG client-side. A roughly 50-line Python backend extends FastAPI, exposes a standard HTML route, and defines an @app.api("remove_background") endpoint; @spaces.GPU handles ZeroGPU allocation while Gradio supplies queuing, concurrency control, SSE streaming, and gradio_client compatibility. The frontend calls the endpoint through the Gradio JavaScript client, so image-processing requests pass through the queue while rendering and compositing remain in the browser. The result is a custom frontend that retains Spaces hosting and Gradio infrastructure.
Agentic coding at ClickHouse
Industry perspectives on AI coding agents remain divided between dismissal and excessive automation claims. ClickHouse integrated agentic coding across internal workflows, progressing from simple boilerplate scripts to active backend C++ development. Engineers primarily use command-line agents such as Claude Code with Claude Opus 4.5 and Codex CLI alongside observability platforms like Langfuse and Librechat. While agents effectively investigate continuous integration logs, generate internal dashboards, and assist with small C++ tasks, they lack architectural competence and require rigorous code review. ClickHouse treats agents as pair programmers for contained tasks, retaining human ownership over product strategy, attention to detail, and core code quality.
We taught ClickStack to read your logs like a detective novel
HyperDX introduced AI Summarize, a feature designed to translate raw logs, traces, and log patterns into narrative summaries directly within its observability interface. Instead of relying on backend large language models, the tool runs entirely client-side using OpenTelemetry-aware data extraction combined with hand-written phrase pools. The generation logic inspects Kubernetes and OpenTelemetry attributes—such as service names, endpoints, exception details, and query durations—to tailor tone across detective noir, Shakespearean drama, and nature documentary themes. Because processing occurs inside the browser tab without external API calls, the system avoids third-party latency, token expenses, model hallucinations, and data privacy risks. Users can access summaries across logs, spans, and pattern drawers or re-enable the interface toggle via URL parameter through April 2026.
Headless Screenshot Testing in Thumbtack’s Android Apps
Thumbtack required a faster, emulator-free method to execute over 1,500 Android screenshot tests during continuous integration. Previously, running UI tests on physical devices or emulators proved too slow for code review verification jobs, causing tests to run post-merge and allowing regressions to persist. After evaluating Paparazzi and Google's experimental Compose Preview Testing, the team selected Roborazzi, which runs on top of Robolectric's native graphics engine within a standard JVM. This headless setup accommodates custom Activities, Dagger field injection, and legacy View systems while executing alongside standard unit tests. Moving screenshot tests to the unit test level enabled gating merges on passing tests, colocating tests inside feature modules, and increasing total CI job runtime by only twenty percent.
From Custom to Open: Scalable Network Probing and HTTP/3 Readiness with Prometheus
Slack encountered a lack of client-side observability when introducing HTTP/3 at its edge. Because HTTP/3 runs over QUIC on UDP instead of TCP, neither existing commercial SaaS tools nor the internal Prometheus Blackbox Exporter could probe the new endpoints. To address this gap, an engineering intern integrated the open-source quic-go library into Prometheus Blackbox Exporter to build a native, configurable HTTP/3 probe. Slack deployed an internal implementation using the upstream functionality while waiting for pull request reviews, ultimately landing the contribution upstream. The resulting setup provides a unified view of HTTP/1.1, HTTP/2, and HTTP/3 metrics within Grafana, enabling reliable alerting and faster correlation with other telemetry.
Granite 4.0 3B Vision: Compact Multimodal Intelligence for Enterprise Documents
Granite 4.0 3B Vision is a compact vision-language model designed to extract information from enterprise documents, including tables, charts, forms, and semantically meaningful key-value pairs. It combines a million-scale ChartNet dataset with a code-guided synthesis pipeline, DeepStack Injection for routing semantic and high-resolution visual features to different layers, and a LoRA adapter architecture built on Granite 4.0 Micro. The modular design supports standalone image processing, text-only fallback, and integration with Docling for multi-page document pipelines. Reported evaluations show 86.4% on Chart2Summary, leading the evaluated models, second place on Chart2CSV at 62.1%, leading scores on the listed table benchmarks, and 85.5% zero-shot exact-match accuracy on VAREX.
Introducing MongoDB Agent Skills and Plugins for Coding Agents
Coding agents often default to relational patterns, resulting in over-normalized schemas, underused compound indexes, and misconfigured search indexes in production MongoDB environments. To address these pitfalls, MongoDB launched official MongoDB Agent Skills alongside integrated plugins for tools including Claude Code, Cursor, Gemini CLI, and VS Code. These skills provide agents with discoverable instructions, schema design heuristics, indexing strategies, and operational safeguards across the development lifecycle. The skills operate alongside the MongoDB MCP Server, which handles database connectivity, authentication, and permission-based governance. Teams can install these bundled extensions directly or manage them via the Vercel Skills CLI to embed MongoDB architectural standards into automated workflows.
Training mRNA Language Models Across 25 Species for $165
OpenMed describes an end-to-end protein engineering pipeline that connects ESMFold structure prediction, ProteinMPNN sequence design, and custom mRNA codon optimization. The codon-language-model study compares CodonBERT, ModernBERT, and several CodonRoBERTa variants using 250,000 E. coli coding sequences, a 69-token codon vocabulary, masked-language modeling, and training on four 80GB A100 GPUs. CodonRoBERTa-large-v2 achieved perplexity 4.10, CAI Spearman correlation 0.404, and 7.7% synonymous recovery, while ModernBERT-base reached perplexity 26.24 and correlation 0.070. The work then trained a multi-species suite on 381,283 sequences spanning 25 species, producing four production models in 55 GPU-hours and documenting planned extensions such as larger training data and additional prediction heads.
Australia and Anthropic sign AI safety MOU
Anthropic and the Australian government signed a Memorandum of Understanding to cooperate on AI safety research and support Australia’s National AI Plan. The agreement centers on collaboration with Australia’s AI Safety Institute, including sharing findings about emerging model capabilities and risks, participating in joint safety and security evaluations, and conducting research with Australian academic institutions. Anthropic will also share Anthropic Economic Index data to track AI adoption, economic effects, and workforce implications, initially covering natural resources, agriculture, healthcare, and financial services. Separately, AUD$3 million in Claude API credits will support four Australian institutions working on clinical genomics, precision medicine, pediatric research, and computing education. The company is also exploring data-center and energy investments, launching a startup credit program, and preparing to open a Sydney office.
Boost Training Goodput: How Continuous Checkpointing Optimizes Reliability in Orbax and MaxText
Orbax and MaxText introduce continuous checkpointing to balance training reliability and performance when hardware failures or preemption events interrupt long-running jobs. Fixed checkpoint intervals create a trade-off: infrequent saves risk substantial lost work, while frequent saves can block or bottleneck training, particularly over unstable networks. MaxText enables asynchronous checkpointing that attempts a new save only after the previous background request is fulfilled; its continuous mode overrides checkpoint_period, and max_num_checkpoints_to_keep can limit storage use. A llama-3.1-70B benchmark on two v5p-128 slices found smaller P50 checkpoint intervals, but average training step time increased because of more frequent device-to-host transfers; the analysis estimated resource conservation against mean-time-between-failure assumptions. Orbax also supports minimum intervals and customizable preservation policies, while co-locating storage with training is recommended because cross-metro bandwidth can reduce checkpointing speed and reliability.
TRL v1.0: Post-Training Library Built to Move with the Field
TRL v1.0 marks a transition from research codebase to a library intended for production systems, with clearer stability expectations. It implements more than 75 post-training methods, but its design prioritizes trying, comparing, and using them over coverage alone. The library responds to a field in which PPO, preference optimization methods such as DPO, ORPO, and KTO, and RLVR-style approaches such as GRPO repeatedly changed which components of the training stack appear essential. TRL addresses that volatility with a stable core under semantic versioning, an experimental surface for faster-moving methods, and deliberately limited abstractions that favor explicit implementations over generic hierarchies. v1.0 also follows distributed breaking changes across the 0.x releases, so migration from the last 0.x version is described as minimal, while future work targets asynchronous GRPO, scaling, and actionable training diagnostics.
Top topics
- Architecture834
- Performance816
- LLMs660
- Machine Learning622
- Open Source563
- AI Agents553
- AI470
- Developer Experience448
Weekly digest
The best engineering reads, once a week.