Loading…
Latest reads
Every post, summarised
Grounded summaries of engineering blogs, newest first. Each links to the original.
- 4,699 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 Cursor Router chooses the right model for the task
Cursor Router selects models for developer turns using production behavior rather than benchmark scores, balancing user satisfaction and cost. Its routing system first uses Compass, a continuous 0-to-1 complexity predictor trained on follow-up behavior, to keep simple turns on a price-efficient model or upgrade complex ones. Then a taxonomy of domains, tasks, and modifiers compares frontier-model strengths, while an optimizer applies a one-sided 75% uplift threshold and each mode's cost budget. The dataset contains hundreds of thousands of live Cursor turns, preserving conversation signals, model-switching effects, inferred performance from subsequent user actions, and token-based costs including cache misses. Current results report Auto Intelligence at above Fable-level satisfaction with 68% lower cost, and Auto Balance outperforming Opus 4.8 at 41% lower cost while raising satisfaction by 3%.
How we built an automated debugging workflow at Sentry
Sentry automated its internal debugging and issue triage workflows to handle bugs and pull requests generated by AI agents. The company deployed Seer, an AI-powered debugging agent, across several repositories to determine root causes, create fix plans, and open pull requests. To manage the resulting influx of pull requests, Sentry built an automated Slack routine powered by Claude that periodically scans for unaddressed pull requests opened within four hours. The routine asks Seer to identify relevant reviewers using GitHub commit history, verifies pull request status, and tags engineers to request merges, closures, or feedback. Early metrics show a 21 percent increase in pull request action rates, a 13 percent increase in 48-hour response rates, and a 12.5 percent rise in pull requests closed without merging due to duplicate or expanded fixes.
DeepSeek-V4 Flash 0731 vs GPT-5.6 Luna on DeepSWE: Cost and Coding
On the DeepSWE coding benchmark, GPT-5.6 Luna achieves 67.2% pass@1 versus DeepSeek-V4 Flash’s 53.3%, a 14-point gap. DeepSeek costs $0.10 per task, roughly six times less than Luna’s $0.61, delivering 532 solves per $100 versus Luna’s 110. A cascade that runs DeepSeek first and escalates only on failure solves 78.9% of tasks at $0.385 each, beating Luna’s solo accuracy and cutting cost by 37%. DeepSeek fails more gracefully (9% broken tests vs 15%) but is slower and dramatically weaker on JavaScript (35% vs 60%). Luna dominates seven of eight task domains, while DeepSeek leads only on query and config languages, illustrating how inexpensive models can combine with flagships for superior price-performance.
How to Show Your Logo in Apple Mail
Displaying a company logo in Apple Mail can improve inbox recognition, trust, and visual distinctiveness, while cited studies associate logos with higher brand recall, open rates, purchase likelihood, and email confidence. The article presents Apple Branded Mail, a proprietary format shipped in iOS 18.2 in 2024, as a free alternative to inconsistent provider settings and the paid certificates and trademarks sometimes needed for broad BIMI coverage. Setup begins with a DMARC policy of p=quarantine or p=reject, then uses Apple Business Connect to add a brand, upload a PNG, HEIF, or JPEG logo at least 1024 × 1024 pixels, select domains, and verify identity. Verification can take up to seven business days; root-domain branding covers unconfigured subdomains, while brands needing broader client coverage are directed to BIMI.
Agent Plugins package your skills, tools, and more
Agent Plugins 1.0.0 is an open, vendor-neutral specification for packaging Agent Skills and MCP servers into portable plugins, addressing incompatible wrappers that force authors to maintain client-specific copies. It defines a plugin as a directory with plugin.json, skills/ subdirectories, mcp.json, and client-owned reverse-domain extension namespaces. The manifest fixes component locations, requires explicit MCP server types for stdio, Streamable HTTP, or legacy HTTP+SSE, and lets skills continue loading when an MCP entry fails. Version 1 deliberately excludes installation, distribution, permissions, sandboxing, trust verification, and user experience, leaving those concerns to clients and related ecosystem layers such as discovery and catalogs. Google is joining the Core Maintainers and says Agents CLI and Data Agent Kit support the format, while more product integrations are expected.
ClickHouse Release 26.7
ClickHouse 26.7 is a monthly release with 61 new features, 112 performance optimizations, and 329 bug fixes, including improvements to aggregation, joins, vector search, phrase search, query analysis, and URL access. Its combination of optimize_aggregation_in_order and the new optimize_aggregation_in_order_limit turns matching GROUP BY, ORDER BY, and LIMIT queries into low-memory streaming pipelines that can stop reading after N groups. In a TPC-H scale-factor-100 test, the optimized query ran 313× faster, processed 339× fewer rows, and used 592× less peak memory. The release also adds build-side JOIN pruning, LLVM JIT compilation for supported regular expressions, and macOS improvements including faster allocation-heavy workloads, concurrent remote-replica reads, SSD cache dictionaries, FileLog, and STREAM support.
How Physical Intelligence unified its robotics data stack with Postgres managed by ClickHouse
Physical Intelligence, which builds robotics foundation models, needed a data stack that could support both transactional operations and high-cardinality analytical workloads. Its previous single RDS Postgres instance could not handle those demands, so the team migrated to ClickHouse Cloud and adopted Postgres managed by ClickHouse for OLTP alongside ClickHouse for OLAP. Postgres is automatically replicated into ClickHouse through ClickPipes, while materialized views, full-text indices, and a query agent help researchers search across the company’s data. The resulting stack supports faster, easier exploration as data volume grows from the 10 billion to 100 billion row range, and also handles telemetry through ClickHouse with ClickStack and Grafana.
Baseten on Hugging Face Inference Providers 🔥
Baseten is now a supported Inference Provider on the Hugging Face Hub, extending serverless inference on model pages and through the Hugging Face Python and JavaScript SDKs. As part of the initial integration, it supports conversational and text-generation tasks, including access to open-weight models such as Kimi K3, latest DeepSeek V4 Flash, and GLM-5.2, with additional tasks planned to roll out soon. Developers can use a custom provider key for direct calls or authenticate with a Hugging Face token so requests are routed to Baseten; model pages and account settings support provider compatibility and preference ordering. Python and JavaScript OpenAI-compatible examples use the Hugging Face router, while integrated agent harnesses can access Baseten-hosted models without extra glue code.
From User Sequences to Scaling Laws: A Multi-Stage Architecture for Meta’s Ads Ranking
Meta’s ads ranking systems process millions of candidates per second under millisecond latency budgets, creating tension between richer sequence models and efficient serving. The post presents a multi-stage architecture that moves long-history user modeling into asynchronous offline transformers, caches user-level embeddings, and combines them with fresh user and ad signals in an online ranking model. Dense tokenization places sparse features and behavioral sequences in one vocabulary, while target-aware multi-head attention learns higher-order interactions between historical behavior and the candidate ad. On real-world ads traffic, performance follows a log-linear relationship with compute across depth, enrichment, width, and sequence length, with balanced growth and diverse sequences identified as scaling levers. Together with broader model innovations, the approach contributed to a cumulative 6% conversion lift on Instagram, 3% on Facebook, and 3.5% lift in Facebook ad clicks, and serves as a core component of GEM.
How Salesforce Eliminated Single-Region Risk and Reduced Downtime Blast Radius at 4B Metrics/Min
Salesforce redesigned its internal observability platform, Argus, to eliminate single-region failure risks while ingesting approximately 4 billion metrics per minute. Operating out of a single AWS region previously created a global blast radius and generated high cross-region data transfer costs. To mitigate these risks without the prohibitive expense of full multi-region replication, the engineering team implemented a geo-local architecture that processes and stores telemetry closer to its origin across production geographies. A newly introduced federation query layer relies on Elasticsearch mappings to route queries selectively to relevant regional clusters, avoiding broad fan-out. The platform also adopted metadata caching for wildcard queries and implemented HTTP 206 partial-response handling to maintain visibility when individual regions experience downtime.
Rider insights: How NJ Transit’s Access Link Riders’ Choice Pilot redefines mobility freedom
NJ TRANSIT’s Access Link Riders’ Choice (ALRC) Pilot 2.0 gives eligible paratransit customers on-demand Uber rides across five New Jersey service regions, with up to $20 off 10 trips monthly for work, appointments, and leisure. Launched in 2024, the pilot addresses traditional paratransit constraints such as advance booking and fixed time windows; Uber data as of July 30, 2026, records more than 269,000 rides. In an Easterseals webinar, rider Wanda Williford and caregiver Sharice Frazile described same-day availability as life-changing, reducing scheduling stress and enabling unexpected medical visits and other travel. Their accounts portray flexibility, real-time ride access, and spontaneity as practical changes to daily routines for riders and caregivers, while Williford calls the program a “game changer” for New Jersey paratransit.
What is WAL backpressure, and why does ClickHouse Managed Postgres need it?
Postgres writes every change to the write-ahead log (WAL), and Managed Postgres uploads completed segments to object storage for point-in-time recovery; if uploads lag, retained WAL can fill the disk and cause a PANIC shutdown. ClickHouse Managed Postgres counters this with a data-plane systemd timer that checks pending segments every 15 seconds and uses cgroup v2 I/O controls to throttle client write bandwidth at 80%, 50%, or 20% of provisioned throughput as backlog thresholds rise. The drain path—including the archiver, checkpointer, and related processes—is placed in an immune cgroup, while client backends are throttled and reads remain uncapped. In a test with a 500 MB/s gp3 volume and an archive limited to 4 MB/s, the backlog reached 3,433 segments and 53 GiB before recovery. Once the limit was removed, the archiver cleared the queue, the throttle released automatically, and the disk stayed below 33% full.
amazon.science
34 Amazon Research Awards Build on Trainium recipients announced
Amazon announced 34 recipients from 30 universities for the Fall 2025 Build on Trainium: Responsible AI call for proposals. Part of a 110 million dollar credit initiative, the program funds academic research across new model architectures, machine learning libraries, optimizations, and distributed systems running on AWS Trainium hardware. The funded projects center on five key areas: AI safety and alignment, multilingual language models, representation engineering, sustainability with small language models, and deep learning for synthetic data generation. Awardees receive AWS Promotional Credits, access to more than 700 Amazon public datasets, assigned Amazon research contacts, and dedicated technical training resources. Highlighted efforts include parallelization strategies for trillion-parameter mixture-of-experts models scaling across 1,024 Trainium chips and inference optimization frameworks designed to enhance token efficiency.
Your OTel spans, our errors: A Sentry love story in one trace
Pointing a standalone OpenTelemetry SDK directly at Sentry ingests spans over standard OTLP, but the spans remain disconnected from Sentry error data. Because the separate SDKs share a process without sharing context, Sentry errors lack active OpenTelemetry trace and span IDs. Sentry introduced OtlpIntegration across backend SDKs including Python, Ruby, Node.js, Go, PHP, .NET, and Java to bridge this gap. The integration reads the active OpenTelemetry context, attaches trace and span identifiers to Sentry events, and configures ingestion. This decoupled model replaces the deeply integrated POTel strategy across most languages, except Java, removing mandatory OpenTelemetry dependencies while maintaining unified trace waterfalls.
Explorers, exploiters, and the myth of the 100x engineer
Engineering leadership often responds to rapid AI productivity gains by attempting to isolate and replicate the traits of individual outlier engineers. Vivek Raghunathan, SVP of engineering at Snowflake, frames this dynamic using reinforcement learning's explore-versus-exploit split, where roughly five percent of an organization fearlessly tests new AI tools while ninety-five percent prioritize execution on paved paths. Instead of treating these categories as fixed personality types or relying on external hiring, organizations succeed by viewing them as a continuum. Effective leadership establishes structured learning time, AI communities of practice, and direct mentorship to capture explorer discoveries and elevate the median engineer. Measuring progress requires tracking organizational movement along this capability scale rather than celebrating isolated high-performing individuals.
Postgres Changes gets AND filters, new operators, and column selection
Postgres Changes now supports filters composed across multiple columns, additional operators, and column selection for subscriptions over Postgres tables. It can receive insert, update, and delete events over WebSockets, while comma-separated conditions or the postgresChangesFilter builder combine conditions with AND. New operators include like, ilike, is, match, imatch, isdistinct, and negated forms, while selected payloads always include the table's primary key. Column selection is opt-in, requires @supabase/supabase-js 2.109.0 or newer, and columns must be selectable by the subscribing role; Row Level Security still applies. These changes can lower billable usage because rejected events avoid messages per subscriber and trimmed payloads send fewer egress bytes, although DELETE events carry only the primary key and OR composition is unsupported.
A beginner’s guide to Uber vouchers for riders
Uber connects riders with drivers through a mobile application, while Uber Vouchers lets transit agencies, universities, and colleges subsidize eligible rides. A voucher added to a rider’s account generally covers trip costs up to a stated amount; any excess is charged to the default payment method, while tips remain optional and are charged separately. The guide explains how to download the app, create an account with matching contact information, link payment, track and verify the arriving vehicle, complete a trip, rate the driver, and access in-app Help. It also covers requesting rides by text-capable phone through 833-USE-UBER (833-873-8237) from 4am to 10pm ET, service-animal access, lost items, and the in-app Safety Toolkit.
Scaling AI Agent Infrastructure with the MCP Stateless updates
The 2026-07-28 Model Context Protocol specification release candidate addresses production bottlenecks caused by the earlier session-oriented HTTP model, including session pinning, sticky routing, lost state, and Redis-backed infrastructure. It removes the initialize/initialized handshake and Mcp-Session-Id, carrying protocol version, client information, and capabilities in _meta on each request while exposing method and name through HTTP headers. The update also introduces ttlMs/cacheScope caching, Multi Round-Trip Requests for elicitations and task progress, issuer and resource-indicator verification, and full JSON Schema 2020-12 support. The post says the stateless core enables ordinary load balancing, serverless deployment, and transparent failover, while SDK beta releases for TypeScript, Python, Go, and C# support migration.
Sidekick's continual learning loop
Frontier models can accelerate an AI product launch, but their serving cost and latency can become impractical as usage grows, while their weights do not learn from production failures. Shopify’s proposed flywheel captures low-quality conversations, defines quality with a rubric and calibrated judges, improves the surrounding harness through autoresearch, and then converts repaired trajectories into supervised fine-tuning and GRPO training signal. A panel of frontier reasoning models critiques failures, an arbiter produces repair instructions, and unresolved cases go to expert annotators; gist compression then shortens the agent’s static prompt without measured judge-quality loss. In production, the GraphQL agent serves up to 2,000 requests per minute and reportedly achieves a 96% serving-cost reduction versus an estimated frontier-model cost, with lower latency and higher throughput. The approach is presented as a continual-learning system that turns production experience into model-weight updates.
Access On Demand pilot with RTD’s Access-a-Ride
RTD’s Access-on-Demand pilot gives eligible Access-a-Ride customers subsidized, curb-to-curb Uber trips that can be requested the same day instead of scheduled in advance. The program complements rather than replaces Access-a-Ride, offers up to 60 subsidized trips per month across participating providers, and operates throughout RTD’s service area during Access-a-Ride hours, except 1:30 a.m. to 3:30 a.m. Riders pay a $4.50 standard fare or $2.25 LiVE fare, while RTD subsidizes up to $20 per eligible trip; riders cover amounts beyond that limit. To use the benefit, customers need active certification, must complete RTD enrollment, and activate an Uber voucher linked to their enrolled account, which applies automatically to eligible trips.
Top topics
- Architecture834
- Performance816
- LLMs665
- Machine Learning622
- Open Source566
- AI Agents557
- AI475
- Developer Experience448
Weekly digest
The best engineering reads, once a week.