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
Updates to Consumer Terms and Privacy Policy
Anthropic is updating its Consumer Terms and Privacy Policy for Claude Free, Pro, and Max users, including Claude Code sessions tied to those accounts. Users can choose whether their data is used to improve Claude and strengthen safeguards against scams, abuse, and harmful content; the change excludes Commercial Terms services and API use. New users choose during signup, while existing users must decide by October 8, 2025; accepting immediately applies the updates to new or resumed chats and coding sessions, and the setting remains changeable in Privacy Settings. For users who opt in, retention will extend to five years, compared with 30 days otherwise, while deleted conversations are excluded from future training; Anthropic says it filters or obfuscates sensitive data and does not sell user data.
First Block with Brendan Foody, Co-founder and CEO of Mercor
First Block features Brendan Foody, co-founder and CEO of Mercor, discussing the company’s path from selling donuts in eighth grade to building an AI-powered recruiting platform. Mercor reportedly grew from $1 million to $100 million in revenue run rate in 11 months, and Foody connects that growth to automated hiring solutions and the AI talent revolution. He argues that founders should pursue problems they cannot stop thinking about, set a high bar for their first 10 hires, and remain patient while shaping culture. The episode also frames fundraising as an input to better user experiences rather than an objective, encourages founders to return quickly to customers, and covers long-term orientation, personal scaling, and finding the right market.
Amazon Nova Models: Now Available on Heroku
Amazon Nova’s nova-pro and nova-lite models are now available through Heroku Managed Inference and Agents, giving applications a managed way to use them. The integration initially supports text-to-text workloads, while multimodal support is planned for a future release; both models support tool use and offer 300,000-token context windows. Nova-lite targets speed, low latency, and high-volume real-time applications at $0.06 per 1M input tokens and $0.24 output, whereas nova-pro targets accuracy and deep reasoning at $0.80 input and $3.20 output. Heroku provides an OpenAI-compatible /v1/chat/completions endpoint for chat and generation, plus /v1/agents/heroku for built-in Heroku tools and custom MCP tools, with provisioning available in US and EU regions through the add-on or CLI.
Engineering stories behind the Medium Daily Digest Algorithm: Part 1
Medium’s Daily Digest recommendations were less relevant than the homepage despite identical story sourcing, ranking model, and features, because digest filtering operated with less email-reporting information and was too aggressive. The “Sent in Opened Digest” filter excluded all 15 stories after an email-open signal, even when a reader saw only the first five; Apple Mail Privacy Protection also caused servers to preload messages, making many open events unreliable. Medium removed that filter and reduced the backoff from seven to four days, accepting more repetition while relying on other filtering rules and organic recommendation changes to diversify digests. An A/B test found that experiment users were 10% more likely to convert to paying members and read 7% more on Medium. The post says these results led to more filter investigations, including Bloom Filters, “Hard vs Soft Filtering,” and cross-digest diversification, while recommending audits of products that depend on email-open tracking.
Building production-ready agentic systems: Lessons from Shopify Sidekick
Shopify’s Sidekick evolved from a simple tool-calling system into an agentic assistant that helps merchants manage stores through natural language, exposing scaling challenges in architecture, evaluation, and training. As its inventory grew beyond 50 tools, the system prompt accumulated conflicting special cases and became difficult to maintain; JIT instructions instead return relevant guidance with tool data. For evaluation, the team uses Ground Truth Sets based on production distributions, human labels from at least three product experts, statistical agreement measures, specialized LLM judges calibrated against humans, and merchant simulation. Judge correlation rose from Cohen’s Kappa 0.02 to 0.61 against a human baseline of 0.69, while syntax validation improved from about 93% to 99% after reward-hacking fixes. GRPO training exposed opt-out, tag, and schema hacking, and the post recommends modularity, procedural plus semantic validation, user simulation, and iterative judge refinement for production agentic systems.
Engineering stories behind the Medium Daily Digest Algorithm: Part 4
Medium’s Daily Digest team investigated complaints that recommendations had become repetitive, especially for power users, after earlier filtering changes. Data showed some stories were sent up to eight times to one user, while affected users saw digest posts repeated twice on average; the team also had to account for Apple Mail Privacy Protection, which prevents reliable open tracking. They therefore tracked digest email IDs when readers clicked a story, marking emails with at least one click as “definitely opened,” and added a recommendation-stack filter that counts prior occurrences in those digests. Testing supported a threshold of one occurrence: complaining users would lose nearly five stories per digest, whereas free users would lose fewer than one. An A/B test produced a statistically significant increase in reading time, with a larger rise among members; the team reported no further complaints and suggested soft filtering and email-position rules as possible refinements.
Build Better Agents With MorphLLM
The post addresses a bottleneck in AI agents that write or edit code: correcting a small change can trigger a full-file rewrite, consuming tokens and causing formatting failures, silent omissions, and unreliable results. MorphLLM’s Morph Fast Apply is presented as a semantic, structure-aware edit merge tool: an agent emits small edit_file snippets with //...existing code... placeholders, while the Apply API merges them against the codebase context before the edited file is written back. The post reports over 4,500 tokens per second, 98% accuracy in about six seconds per file, and about four seconds for files up to 9k tokens, compared with 35 seconds and 86% accuracy for traditional search-and-replace with error corrections. It also states that MorphLLM is API-compatible with OpenAI, Vercel AI SDK, MCP, and OpenRouter, and can run in the cloud, self-hosted, or on-premises.
Real-time vs batch reconciliation: Practical patterns for keeping data in sync
Product data reconciliation across internal databases, ERPs, and platforms such as Stripe becomes harder to maintain as catalogs and system relationships grow, affecting financial reporting, customer experience, and operations. The post compares real-time reconciliation, using tools such as Kafka, Flink, Delta Live Tables, and Debezium for more immediate consistency, with batch approaches built around Spark, AWS Glue, Snowflake, and dbt for periodic, large-scale checks at lower cost and complexity. It also presents validation options including Great Expectations, Apache NiFi, Talend, and Airbyte, and recommends defining data ownership, applying field-specific conflict rules, and routing critical mismatches to manual review. Soft deletes and tombstones preserve deletion history, while observability through synchronization latency, conflict rates, resolution patterns, and regular audits helps assess reconciliation health as product catalogs scale.
From naive webhooks to durable sync: Queue-based and event-driven data reconciliation patterns
Maintaining consistent product data between an internal catalog and Stripe becomes difficult when outages, throttling, high volumes, and changes from both systems create reconciliation failures. The post compares a naive one-way flow, in which a Lambda calls the Stripe API with no persistent synchronization record, with a queue-based design that durably stores changes, controls processing rate, retries failures with exponential backoff, and sends persistent failures to dead-letter queues. For bidirectional synchronization, it contrasts periodic ETL, using extraction, transformation, and loading, with an event-driven model based on Stripe Event Destinations, EventBridge, SQS, and symmetrical events from internal systems. The conclusion is that durable, observable queues and event-driven processing provide buffering, auditability, near-real-time propagation, independent change processing, and more graceful scaling for distributed reconciliation.
An Internship on Ramp’s Forward Deployed Engineering Team
The supplied material consists of the title “An Internship on Ramp’s Forward Deployed Engineering Team” and page markup, but it does not provide narrative details about the internship, the team, responsibilities, technical work, or outcomes. The markup identifies the site as “Ramp Builders,” supplies an RSS feed titled “Ramp Builders Blog,” and includes standard icons, social metadata, and a JavaScript application entry point. It also states that JavaScript must be enabled to run the app, while linking a stylesheet and several favicon assets. Analytics configuration includes Google Analytics and Segment, with a Ramp-hosted Segment endpoint and Sprig (Actions) disabled; these details describe the page infrastructure rather than the internship.
Bringing the Cursor Agent to Linear
Cursor has introduced a Linear integration that lets teams delegate issues to @Cursor by assignment or comment mention, then runs a background agent with issue details, comments, and linked references as context. Teams can also define triaging rules that automatically assign selected tasks and review agent diffs and pull requests from Linear. After connecting a Linear team through Cursor’s integrations page, users assign an issue or mention @Cursor; the agent analyzes it and notifies users when the work is complete and a pull request is opened. Cursor says its team uses the workflow for thousands of customer-feedback, feature, and bug tasks, including a product-bug and feedback triage rule. According to the post, the agent now drafts pull requests for thousands of issues and can one-shot many of them.
Developing nuclear safeguards for AI through public-private partnership
AI models are increasingly capable, raising concern that they could provide dangerous nuclear-technical knowledge that threatens national security. Nuclear information is especially difficult for a private company to evaluate alone, motivating a partnership with the U.S. Department of Energy’s National Nuclear Security Administration (NNSA). Together with NNSA and DOE national laboratories, the company co-developed a classifier that distinguishes concerning from benign nuclear-related conversations, achieving 96% accuracy in preliminary testing. The classifier has been deployed on Claude traffic within a broader system for identifying model misuse, and early deployment data suggests it works well on real conversations; the approach will be shared with the Frontier Model Forum as a potential blueprint for similar safeguards developed with NNSA.
Keeping product data aligned with Stripe as your systems scale
Keeping internal product data aligned with Stripe becomes an architectural challenge as catalogs grow from hundreds to millions of items. The problem spans independently operated systems with different data models, API constraints, rate limits, availability guarantees, and consistency models, while products may change faster than periodic reconciliation can complete. Stripe's payment-focused model covers identifiers, descriptions, prices, currencies, billing terms, and subscription behavior, whereas internal systems may contain categories, inventory, rich content, technical specifications, media, and sensitive cost data. The series examines manual verification, simple scripts, and periodic jobs for smaller catalogs, then points to streaming, change data capture, intelligent partitioning, and conflict resolution for high-scale reconciliation. It concludes that cross-system synchronization requires deliberate mapping and automated architecture because perfect synchronization is difficult to guarantee in eventually consistent distributed systems.
Gating entrances with Stripe and NFC passes
The post presents a gym membership access system that replaces plastic cards, fobs, and physical keys with Apple Wallet and Google Wallet passes scanned by an NFC reader. It uses Stripe Checkout and subscriptions for payment, PassNinja as a single API layer for issuing passes, and a DotOrigin VTAP100 connected to a Raspberry Pi Zero 2W, relay, and door strike for entry control. After Checkout, the website verifies that the Stripe subscription is active, creates a pass whose nfc-message contains the Stripe subscription ID, and redirects the member to add it to a wallet. A Linux gate process reads that identifier from the reader's virtual serial port, retrieves the subscription from Stripe, and activates the relay for five seconds when the status is active; otherwise it denies access.
Heroku AI Expands Model Offering with OpenAI’s gpt-oss-120b
Heroku has made OpenAI’s open-weight gpt-oss-120b available through Heroku Managed Inference and Agents, giving developers an OpenAI-compatible chat completions API for compatible SDKs and frameworks. The model uses a Mixture-of-Experts architecture with 117 billion total parameters, activating 5.1 billion per token and enabling operation on a single 80GB GPU; its Apache 2.0 license permits inspection and fine-tuning. It is designed for agentic workflows involving instruction following, function calling, web searches, and Python execution, while reported benchmarks place it near or above several open-weight models and sometimes OpenAI o4-mini. The add-on provides deployment, secure application access, agent actions, monitoring, and scaling without infrastructure management, with pricing of $0.15 per million input tokens and $0.60 per million output tokens.
How to Build a Secure iOS App with MFA
This tutorial shows how to add secure authentication and backend authorization to an iOS app using Okta’s BrowserSignIn module from the Okta Swift SDK. It registers a native OIDC client with Authorization Code Flow and PKCE, configures redirect URIs and authorization-server policies, and recommends phishing-resistant MFA factors such as Okta Verify with biometrics and FIDO2 with WebAuthn. The implementation uses Swift Package Manager, an Okta.plist configuration file, a testable AuthService protocol, and SwiftUI views for login, token handling, user information, and authenticated state. It then demonstrates authorized API requests by attaching an access token to a URLRequest, calling a backend endpoint, decoding its response, and displaying a returned message, concluding that OAuth 2.0 and JWT validation provide integrated authorization verification.
Context engineering case studies: Etsy-specific question answering
The post examines whether prompt engineering can provide truthful Etsy-specific answers for AI-assisted onboarding without fine-tuning an LLM, using Travel & Entertainment (T&E) policy questions and an external Etsy community-forum use case. For the internal pilot, T&E documents were indexed, embedding-based search retrieved relevant sections, and those sections were added to each question; a manually curated 40-question evaluation found approximately 86% correct or satisfactory answers, with 14% wrong or misleading. Explicit uncertainty instructions sometimes avoided hallucinations but also missed available answers, while asking for an explanation or chain-of-thought reasoning could improve results, though complex reasoning required careful setup. The experiments conclude that prompt engineering is a practical alternative to parameter-based fine-tuning, but requesting specific source snippets remains important for flagging potential hallucinations.
Building Data-Aware AI Applications with Heroku AI and LlamaIndex
Organizations building AI applications over private data must connect large language models with proprietary datasets without managing every infrastructure component. The post presents Heroku Managed Inference and Agents with LlamaIndex as a stack for retrieval-augmented generation (RAG), combining LlamaHub connectors, indexing, retrieval, and managed model services. The implementation provisions Heroku Postgres with pgvector, Claude 4 Sonnet for inference, and Cohere Embed Multilingual embeddings, then uses Python and LlamaIndex to load documents, store 1,024-dimensional vectors, and query them through an OpenAI-compatible embedding endpoint. PGVectorStore supplies the vector store, while Heroku environment variables provide service credentials and connection details. The result is a deployable pattern for querying private documents, with examples spanning financial analysis, insurance claims, customer support, and internal knowledge.
Data mesh at Grab part I: Building trust through certification
Grab began its Signals Marketplace data mesh journey in 2024 because its expanding, varied data ecosystem had ownership gaps, duplicated pipelines, central Data Engineering bottlenecks, broken downstream dependencies, inconsistent definitions, and uneven practitioner expertise. The decentralised model treats data as a product owned by domain teams, with Business Data Owners and Technical Data Owners accountable for upkeep, documentation, usability, and SLOs. Certification formalises expectations through data contracts covering schema, freshness, completeness, retention, change notice, and communication, while automated Data Production Incidents flag breached reliability guarantees for investigation. Using the percentage of queries hitting certified assets as its north star metric, Grab reports 75% of queries targeting certified assets, a 400% year-over-year increase in deprecated tables, and a 58% reduction in P80 datasets.
Trust Calibration for AI Software Builders
Trust calibration means aligning users’ trust in an AI product with its actual capabilities, avoiding both dangerous over-reliance and unnecessary rejection. Drawing on a 2023 review of more than 1,000 papers on trust in human/automated systems, the article distinguishes cooperative tools, where users accept or reject suggestions, from delegative systems, which replace human action and need stronger boundaries, fallbacks, and capability communication. It recommends calibrating expectations before and during interaction through capability-focused onboarding, context-sensitive feedback, visible suggestion boundaries, confirmation friction for high-stakes actions, and layered performance and process information; post-interaction reflection can still support learning. The article concludes that adaptive calibration is more effective than static approaches, anthropomorphic design can create unwarranted trust, explanations can overload users, and builders need measurements beyond satisfaction or adoption.
Top topics
- Architecture834
- Performance816
- LLMs660
- Machine Learning622
- Open Source563
- AI Agents553
- AI470
- Developer Experience448
Weekly digest
The best engineering reads, once a week.