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
Waypoint-1.5: Higher-Fidelity Interactive Worlds for Everyday GPUs
Overworld’s Waypoint-1.5 is a real-time video world model aimed at making interactive generative worlds usable on consumer hardware rather than limited to datacenter-scale compute. It offers two tiers: a 720p model for higher-performance desktop GPUs, capable of up to 60 FPS on RTX 3090 through 5090 hardware, and a 360p model for broader deployment, including gaming laptops and soon Apple Silicon Macs. Training used nearly 100 times more data than Waypoint-1, while more efficient video-modeling techniques reduce redundant computation across frames to support coherent motion and responsive exploration. Users can run it locally through the Biome desktop client, try it in a browser via Overworld Stream, or use the World Engine inference library and third-party clients. The release presents improved visual fidelity, broader local hardware support, and continued progress toward real-time interactive world simulation.
Multimodal Embedding & Reranker Models with Sentence Transformers
Sentence Transformers v5.4 adds multimodal embedding and reranking support for text, images, audio, and video through its Python API. Embedding models map modalities into a shared space, while multimodal Cross Encoder rerankers score relevance for mixed-modality pairs, enabling cross-modal similarity, retrieval, and multimodal RAG workflows. The guide shows model loading, URL-based image encoding, encode_query() and encode_document() prompt handling, similarity calculation, and ranking, with a Qwen3-VL-Embedding-2B example producing 2,048-dimensional image vectors and matching text queries to images. It notes that cross-modal scores can be lower because of the modality gap, while relative ordering can still support retrieval, and that rerankers generally offer higher quality at the cost of slower per-pair processing; installation extras and GPU memory requirements vary by modality and model.
Selective Test Execution at Stripe: Fast CI for a 50M-line Ruby monorepo
Stripe’s Ruby monorepo contains about 50 million lines, 100,000 test files, and roughly 1.2 million test units, making sequential execution impractical at an estimated four months. To keep CI within minutes, Selective Test Execution (STE) runs tests whose recorded file accesses intersect files changed since a suitable baseline, averaging about 5% of the suite and under 10% of always-run compute. A dynamically linked C++ library loaded with LD_PRELOAD intercepts file opens, associates them with hierarchical test scopes, and propagates tracing to child processes; logs are aggregated into a selection index outside the syscall hot path. Guardrails cover tests affected by directory discovery, previously failing tests, selective linting, generated files, and reproducible database-backed baselines ordered by Monotonic Revision IDs. The system is presented as a way to preserve safety and affordability while scaling CI for a highly dynamic language.
ClickHouse at FOSDEM 2026
The ClickHouse community and development team participated in FOSDEM 2026 in Brussels, hosting a community dinner and delivering technical presentations across multiple developer tracks. Presentations covered a range of engineering topics, including building geospatial analytical applications with Leaflet and ClickHouse, as well as integrating LLVM XRay to dynamically hot-patch production logging without redeployments. Additional ClickHouse sessions addressed the engineering challenges of integrating Rust libraries into large C++ codebases, optimizing low-level memcpy routines, and implementing specialized inverted indexes for large-scale text processing. Contributed community sessions detailed workflows for upstream patches in MariaDB and Postgres, alongside WebShield, an open-source Python daemon that streams web accelerator access logs into ClickHouse for bot analysis.
The three villains to agentic observability: retention, sampling and rollups
Observability practices increasingly depend on data sampling, short retention windows, and rollups to control storage costs. While human operators historically compensated for these compromises with intuition, AI agents lack implicit operational context and require full-fidelity telemetry to diagnose issues and justify conclusions. Head sampling drops unobserved spans early, tail sampling risks missing implicit non-error anomalies, and aggressive log expiration prevents long-term cyclical trend analysis. Addressing these constraints requires decoupling storage and compute via high-compression object storage and dedicated compute pools. Deploying real-time analytical databases such as ClickHouse enables high-concurrency, low-latency SQL querying, log clustering, and structured in-database summarization necessary for automated reasoning loops.
Bugbot now self-improves with learned rules
At its July 2025 beta exit, Bugbot resolved 52% of identified bugs by the relevant PR merge; that rate is now nearing 80%. In the public-repository comparison, Cursor Bugbot reached 78.13% across 50,310 analyzed PRs, versus Greptile’s 63.49%; an LLM judge checked whether comments were addressed by merge. It now turns reactions and replies to Bugbot comments, along with human-reviewer comments, into live feedback rather than relying exclusively on offline experiments. Bugbot converts those signals into candidate rules, evaluates them on incoming PRs, promotes candidates to active status as signal accumulates, and can disable rules receiving consistent negative feedback. More than 110,000 repositories have enabled learning, generating over 44,000 rules, which users can edit or delete in the Cursor Dashboard.
Custom OIDC Providers for Supabase Auth
Supabase has launched Custom OIDC Providers, enabling projects to connect standards-compliant external OpenID Connect identity providers beyond its built-in social options. Configuration requires an issuer URL, client credentials, and scopes; Supabase automatically resolves the discovery document, endpoints, and JWKS, and verifies ID tokens against the provider’s JWKS. After enablement, users use the standard OAuth sign-in flow through existing JavaScript, Flutter, or Swift clients, while the auth server handles redirects, code exchange, profile retrieval, and user creation or linking. Custom providers include default PKCE, authorization-parameter support, multi-platform audience validation through acceptable_client_ids, and an email_optional setting for providers that do not return email addresses. They can be created, listed, updated, or deleted via the Dashboard or Admin API, with a limit of three per project.
Safetensors is Joining the PyTorch Foundation
Safetensors has joined the PyTorch Foundation as a foundation-hosted project under the Linux Foundation, alongside DeepSpeed, Helion, Ray, vLLM, and PyTorch. It began as a Hugging Face project addressing the risk of arbitrary code execution in pickle-based model-weight formats. Its format stores tensor metadata in a JSON header capped at 100MB, followed by raw tensor data, and supports zero-copy mapping and lazy loading. The move places the trademark, repository, and governance with the Linux Foundation, while Luc and Daniel remain on the Technical Steering Committee and lead day-to-day work. Existing users get the same format, APIs, and Hub integration without breaking changes, while contributors gain a documented maintainer path; planned work includes PyTorch core serialization, accelerator-aware loading, parallel loading, and quantization support.
How Eisan made POS analytics faster, cheaper, and more reliable with ClickHouse Cloud
Eisan System Development moved a retail analytics workload from a QlikView-based platform to ClickHouse Cloud after rising ID-POS data volumes caused slower processing, failed analyses, and increasing licensing and server costs. Client records from supermarkets and drugstores support basket analysis, cross analysis, and real-time personalization, with some datasets nearing 200 million records annually. The team compared open-source options including PostgreSQL and YugabyteDB, then tested ClickHouse configurations, including a 30-core, 120-GB setup, while accounting for cache memory usage through load testing. In comparison, ClickHouse ran on slightly lower-spec hardware under high-load conditions without observed performance deterioration, and Eisan reports no cases so far where it failed to return a result; the team is now targeting 100 billion records and exploring off-hours processing and scheduled scaling.
MongoDB Predictive Auto-Scaling: An Experiment
MongoDB Atlas initially relied on a reactive auto-scaler that took minutes to adjust instance sizes and only transitioned between adjacent tiers. Because these delays caused transient server over-utilization or unnecessary customer costs, researchers prototyped predictive auto-scaling to forecast demand. The experimental architecture used short-term and long-term forecasters for customer-driven metrics, an estimator using boosted decision trees to project CPU utilization, and a planner to pick the most cost-effective tier. In testing, the prototype kept replica sets closer to target CPU ranges while saving an average of nine cents per replica set hourly. MongoDB subsequently implemented a production version that scales replica sets up proactively before expected spikes while retaining reactive downscaling.
Frontend Engineering at Palantir: Plotlines in Three.js
Palantir’s frontend engineers are building a 3D implementation of Gaia to visualize Foundry Ontology objects with geospatial properties, including ground locations, satellite positions, and paths through space. The system uses Zodiac, a Three.js library backed by Foundry-based APIs: the backend stores, filters, and computes data, while the frontend receives simplified geometry and style information for rendering. Performance techniques include converting latitude and longitude to Earth-fixed coordinates once, batching similarly styled lines into single draw calls, and using instanced meshes plus batched updates for dense satellite scenes; orbital propagation uses TLEs client-side, while heavier analysis remains server-side. The article also describes frame-synchronized animation callbacks that update material properties without rebuilding geometry, while noting that 3D APIs continue to iterate across usability, platform integration, and performance as new data sources and workflows are added.
Finance Trends 2026: Free Trials, BNPL, and Revoked Benefits
Brandwatch Consumer Research examined finance-related online conversations and found that negativity is increasing, with negative mentions up 21% between March 1, 2025, and March 1, 2026. The analysis identifies three major sources of frustration: free trials that convert into subscriptions through unclear pricing or cancellation practices, brands removing or replacing previously free benefits, and buy-now-pay-later (BNPL) services associated with overspending, debt, and financial stress. Mentions of free trials rose 45%, denied-refund mentions in those discussions rose 107%, benefit downgrades or removals rose 19%, and BNPL’s negative sentiment increased almost 600%. Drawing on 609 million conversations from social media, forums, review sites, and blogs over 12 months, the report recommends consent, advance notice, customer education, and financial-literacy support.
ClickHouse Release 26.3
ClickHouse 26.3 delivers twenty-seven new features, forty performance optimizations, and over two hundred bug fixes. Common table expressions can now be materialized into temporary tables via the MATERIALIZED clause when enabled, evaluating subqueries once to reduce query execution times and peak memory usage. The query optimizer expands join reordering beyond inner and left or right joins to include anti, semi, and full joins when table statistics are available. In addition, user-defined functions can run compiled WebAssembly code inside sandboxed Wasmtime runtimes. Storage for the Map data type also introduces hash-based bucketing during background merges, accelerating single-key lookups by two to forty-nine times without slowing initial inserts.
Zero downtime Upgrade: Yelp’s Cassandra 4.x Upgrade Story
Yelp's Database Reliability Engineering team upgraded over a thousand Cassandra nodes from version 3.11 to 4.1 without downtime. The upgrade delivered reliability, observability, and performance improvements, including Java 11 support, configurable guardrails, and fixed incremental repairs. To avoid the high costs, consistency compromises, and data streaming delays of standing up separate datacenters, the team selected an in-place rolling upgrade strategy. The rollout orchestrated staged node updates alongside compatible versions of the Stargate proxy and custom data pipeline connectors without requiring client code modifications. Although Stargate 2.x introduced regressions on range and multi-partition queries that required downgrading to 1.x, the finished migration achieved up to a 58% reduction in p99 latency across key clusters.
Mobility for all: How Uber supports Los Angeles’s vision of universal basic mobility
Los Angeles’s Mobility Wallet pilot provides low-income residents with funds for shared transportation, including Uber rides, as part of Universal Basic Mobility initiatives. Phase I ran from May 2023 to April 2024 and gave 1,000 South Los Angeles residents a monthly $150 stipend on prepaid cards usable for transit, bikes, e-scooters, regional travel, and ride-hailing. Participants reported improved quality of life, citing redirected transportation spending for bills, school supplies, social activities, and savings, along with reduced financial stress and improved mental health. During the phase, participants took more than 7,600 Uber rides to reach fixed transit, appointments, homes after late shifts, or social events. Phase II began in fall 2024 and expanded eligibility to 2,000 participants across LA County, offering up to $1,800 over 12 months and Uber vouchers, while results remain forthcoming.
Frontend Engineering at Palantir: Building a Backend-less Cross-Application API
Palantir’s Application Frameworks team builds the “backend of the frontend”: frameworks that let applications communicate during workflows spanning multiple interfaces. The work addresses environments with denied, disrupted, intermittent, or limited (DDIL) connectivity, where relying on backend services can introduce unreliable latency. The team generalized the Gaia Controller prototype into Shared State, which lets applications define named data models and exchange updates organized by Media Type, and Commands, which exposes arbitrary logic through command IDs and request objects that return responses. Both frameworks use frontend-only browser APIs for client-side interaction, support workflows such as pairing Gaia with Workshop, and are also being used as tools by AIP agents; the post says adoption is growing across Palantir.
How Respan is scaling LLM observability with ClickHouse Cloud
Respan scaled its LLM gateway and observability platform from hundreds of daily requests to 50 million daily events, causing its original Django and Postgres architecture to stall under transaction contention at 50 to 100 requests per second. To overcome Postgres write-ahead logging bottlenecks, the team migrated the ingestion and analytics pipelines to ClickHouse Cloud. Engineers minimized row sizes by truncating large prompt text and writing structured metrics to MergeTree tables, then built incremental minute- and hour-level materialized views alongside scheduled refreshable views for user metadata. Distributed tracing spans are ingested asynchronously with analytical parent-child resolution, utilizing functions like argMax rather than the FINAL modifier to prevent costly table merges. This architectural shift allows Respan to achieve sub-second search across tens of millions of records and support customer dashboards with predictable query performance.
Better MoE model inference with warp decode
MoE inference commonly organizes decode around experts, but the post describes “warp decode,” an output-centric kernel for small-batch generation on NVIDIA Blackwell GPUs. Instead of staging tokens into expert-major buffers, each warp owns one output scalar, streams the required weight rows, accumulates contributions from all eight routed experts in FP32, and uses warp-level reductions; the gate/up and down projections are compressed into two kernels. This removes padding, scatter, combine, activation gather, and per-expert output buffers, while keeping BF16 activations and avoiding intermediate quantization. On B200s, the approach delivered 1.84x higher decode throughput, sustained 3.95 TB/s at batch size 32, and produced outputs 1.4x closer to a full FP32 reference.
Never stop disrupting yourself; introducing the Fin API platform
Fin API platform launch makes the models powering its customer-agent system available through APIs, with contracts starting at $250k annually. It says Fin currently resolves over 2M customer issues weekly across brands and runs on models trained by its AI group. Last week’s Apex announcement is described as a specialized customer-service LLM that beat frontier models from Anthropic and OpenAI in six-month production tests on resolution rate, latency, hallucination rate, and cost. Customers can choose the Fin Agent Platform, Fin Agent API, or direct access to Apex and related models for bespoke, specialized agents. The rationale is customer demand and a broader expectation that software differentiation will shift from features and interfaces toward agents built on differentiated AI.
Improving storage efficiency in Magic Pocket, our immutable blob store
After introducing a direct erasure-coding service, Dropbox encountered severe storage fragmentation in its immutable blob store, Magic Pocket, due to a large number of under-filled volumes. Because their legacy compaction strategy was only designed to top off near-full volumes, the team created new multi-strategy compaction methods to aggressively consolidate sparse volumes. These approaches leverage dynamic programming and streaming pipelines to rapidly reclaim unused disk capacity at exabyte scale.
Top topics
- Architecture834
- Performance816
- LLMs660
- Machine Learning622
- Open Source563
- AI Agents553
- AI470
- Developer Experience448
Weekly digest
The best engineering reads, once a week.