---
title: "Latest reads"
description: "The engineering internet, summarised so you can actually read it."
---

# Latest reads
> The engineering internet, summarised so you can actually read it.

## Articles

### [Welcoming Phil Nash, our new Developer Experience Engineer](https://yomu.fyi/post/welcoming-phil-nash-our-new-developer-experience-engineer.md)
- Company: [Resend](https://yomu.fyi/company/resend.md)
- Published: Jul 30, 2026
- Type: Announcement

Phil Nash is joining Resend as a Developer Experience (DX) Engineer after eight years at Twilio and subsequent work at IBM. He has taught developers through talks, articles, and live coding, and will focus on helping developers become more productive. His path into software began with building an HTML page at 13, continued through maths and computer science studies, and led to front-end development after creating a WordPress blog and writing about what he learned. Nash says he joined because he had used Resend for a personal-site contact form, built a simple TypeScript API client, and wanted to help make email a better experience for developers. He recommends sharing new learning through writing, presentations, videos, or other community contributions.


### [How to use Google microbenchmarks for evaluating TPU performance](https://yomu.fyi/post/how-to-use-google-microbenchmarks-for-evaluating-tpu-performance.md)
- Company: [Google](https://yomu.fyi/company/google.md)
- Author: Junjie Qian, Chi Shuen Lee, Yu-Hsuan (Amy) Lin, Haixiong (Sean) Wang
- Published: Jul 30, 2026
- Type: Explainer

Evaluating TPU performance requires measurements beyond product specifications, because real-world behavior depends on architecture and workload constraints. The suite measures network collectives, compute, HBM bandwidth, host transfers, and Ragged-Paged Attention primitives, using metrics such as GB/s, latency, TFLOPs, MFU, and TTFT-related throughput. Its src and configs directories provide implementations and tunable parameters, while sample Kubernetes recipes connect measurements to TPU architecture and runtime behavior. Roofline analysis uses peak MXU throughput and HBM bandwidth to distinguish compute-, memory-, and network-bound workloads and guide kernel, sharding, or rematerialization choices. In a 110B MoE workload on a 4x4x4 TPU 7x configuration, microbenchmark-guided changes addressed ICI stalls and memory-bound attention, reducing training step time by 21.2%.


### [Meet Stripe's Knowledge AI Platform](https://yomu.fyi/post/meet-stripe-s-knowledge-ai-platform.md)
- Company: [Stripe](https://yomu.fyi/company/stripe.md)
- Author: Anna Mason
- Published: Jul 30, 2026
- Type: Explainer

Stripe built Kai, a knowledge AI platform for non-engineering work that existing no-code and coding agents could not support with Stripe’s required data security and workflow specificity. It serves tasks including account research, incident triage, revenue modeling, compliance reviews, and data-warehouse queries, while adapting to different tools, data sources, outputs, and definitions of completion. Kai is organized into surface-agnostic APIs, AgentStudio for domain teams to build and govern agents, and shared execution environments with per-session Kubernetes sandboxes, multi-tenant virtual filesystems, access controls, and long-session state management. The platform connects to more than 1,000 skills and tools and supports web, Slack, embedded, and custom application experiences. Within two weeks of its April launch, most of Stripe was using Kai; it now has 83% weekly active users, and reported effects include 25,000 annual hours shifted from administrative work and higher sales activity and deal creation among Account Executives.


### [Crowdsourced taxonomy verification: A feedback-driven framework for refining knowledge graph relationships via online search interactions](https://yomu.fyi/post/crowdsourced-taxonomy-verification-a-feedback-driven-framework-for-ref.md)
- Company: [Grab](https://yomu.fyi/company/grab.md)
- Author: Junpeng Niu
- Published: Jul 30, 2026
- Type: Problem & solution

Dynamic catalogs in food delivery and e-commerce make knowledge-graph (KG) maintenance difficult: LLM- or scraper-generated relationships can be semantically wrong, manual review does not scale, and bad edges can degrade query expansion, CTR, and trust. The proposed feedback-driven verification engine treats unverified edges as hypotheses and uses the search interface as a validation environment. It separates verified and candidate edges, injects a limited number of candidates into lower-risk interface positions through an exploration-versus-exploitation policy, and records context-anchored user interactions with weighted positive and negative signals. An offline engine aggregates exposure and engagement into a confidence ratio, promoting candidates above a verification threshold and pruning those below a failure threshold. The framework is presented as a continuous human-in-the-loop method for validating parent-child and sibling relationships at scale; planned extensions add confidence tiers and defenses against spam, bots, and Sybil attacks.


### [NBCUniversal’s Seamless Migration: Unlocking Scalable Analytics with Databricks](https://yomu.fyi/post/nbcuniversal-s-seamless-migration-unlocking-scalable-analytics-with-da.md)
- Company: [Databricks](https://yomu.fyi/company/databricks.md)
- Author: Kevin Hill, Ludwig Kuznia, Anil Joshi, Jay Mehta, Manojit Dan
- Published: Jul 29, 2026
- Type: Problem & solution

NBCUniversal sought to modernize its data infrastructure as growing data volumes, evolving analytics demands, and slot-based compute reservations created cost, contention, and flexibility concerns. Working with EXL and Databricks, it migrated workloads to a Databricks lakehouse architecture that unifies data engineering, analytics, machine learning, and business intelligence. The phased, risk-based program assessed workloads and dependencies, designed workspaces and governance with Unity Catalog, and used four custom accelerators for SQL translation, DAG migration, data transfer, and validation. Job-specific compute, autoscaling, Delta Lake layouts, MLflow, and Lakeflow Jobs supported workload-specific execution and operations. The reported outcome was a 30% reduction in data infrastructure costs, elastic scaling for high-traffic events, unified ML operations, and the onboarding of approximately 300 analysts to Databricks SQL.


### [Quality care is the mission. Finance protects the margin.](https://yomu.fyi/post/quality-care-is-the-mission-finance-protects-the-margin.md)
- Company: [Databricks](https://yomu.fyi/company/databricks.md)
- Author: Aaron Zavora
- Published: Jul 29, 2026
- Type: Explainer

Healthcare finance teams must protect margins amid rising medical costs, denied or underpaid claims, complex payment arrangements, and cash trapped in receivables, while fragmented systems and delayed information increase decision risk. The post presents ontology as a way to preserve the business meaning behind figures, linking numbers to service lines, payers, contracts, and changing context rather than treating accuracy alone as correctness. Databricks Genie is described as a governed, data-smart AI coworker that answers finance questions with sourced figures, traces them to their origins, respects permissions, and keeps a person in the loop. It applies this model to care costs exceeding reimbursement, revenue lost to denials and underpayments, and aging or unbilled receivables, with the intended actions of renegotiating rates, preventing claim errors, accelerating collections, and protecting margin.


### [Manufacturing runs on capital. Finance protects the margin.](https://yomu.fyi/post/manufacturing-runs-on-capital-finance-protects-the-margin.md)
- Company: [Databricks](https://yomu.fyi/company/databricks.md)
- Author: Caitlin Gordon
- Published: Jul 29, 2026
- Type: Explainer

Manufacturing finance must protect margin by keeping capital moving through inventory, receivables, and plant equipment. The post argues that volatile supply chains, shifting demand, rising costs, and increasingly influential agents make that work faster and more complex, while an estimated $1.7T remains trapped in excess working capital across large US companies. It presents ontology as a way to preserve the meaning and context behind figures, including plants, SKUs, customer terms, and changing business conditions. Databricks Genie is described as a data-smart AI coworker whose ontology learns from business systems and questions, produces sourced and governed answers, and helps finance identify trapped inventory cash, aging receivables, and underperforming assets. It readies actions such as releasing inventory, accelerating collections, or redeploying capital, while a person remains responsible for the decision.


### [Energy runs on volatile markets. Finance protects the margin.](https://yomu.fyi/post/energy-runs-on-volatile-markets-finance-protects-the-margin.md)
- Company: [Databricks](https://yomu.fyi/company/databricks.md)
- Author: Caitlin Gordon
- Published: Jul 29, 2026
- Type: Explainer

Energy finance must protect margin as hourly power and fuel prices, shifting forward curves, contract settlements, hedges, and major grid and generation investments make business conditions more complex. The post argues that accurate figures are not necessarily correct unless they retain context about the underlying asset, market, and contract, motivating a live ontology that keeps business meaning current. It presents Databricks Genie as a data-smart AI coworker that learns from enterprise systems and questions, returns sourced answers, preserves permissions, governs AI cost, and shows its work. The proposed uses cover live margin, revenue risk across PPAs and hedges, and funding AI-driven buildout, with a person making trading, dispatch, hedging, or capital decisions. Genie is described as helping finance identify risks early and prepare actions while keeping human owners accountable.


### [Bringing real-time fraud prevention to government benefits](https://yomu.fyi/post/bringing-real-time-fraud-prevention-to-government-benefits.md)
- Company: [Databricks](https://yomu.fyi/company/databricks.md)
- Author: Mike McWhorter, Johnathan Tafoya, Stephen Hutson
- Published: Jul 29, 2026
- Type: Problem & solution

Federal benefits programs face an estimated $233 billion to $521 billion in annual fraud losses and roughly $186 billion in improper payments in fiscal 2025, while pre-payment vetting can delay legitimate aid. The post presents real-time detection as an alternative to “pay and chase,” combining rules engines, machine learning, adaptive and generative AI, entity resolution, behavioral analysis, and graph analytics to score claims before disbursement. It says Databricks can unify these layers and enable cross-agency sharing through OpenSharing, data masking, and Clean Rooms while retaining governance and data ownership. When fraud is detected, the platform can hold payments, add confirmed fraudsters to Do Not Pay, and assemble statute-linked evidence packets with digitally signed audit trails, while designated personnel authorize referrals. The source concludes that this model is ready for implementation at national scale, citing Databricks’ stated processing of 160 billion card payments annually in milliseconds.


### [Agents for production lines: Trusted decisions in real time](https://yomu.fyi/post/agents-for-production-lines-trusted-decisions-in-real-time.md)
- Company: [Databricks](https://yomu.fyi/company/databricks.md)
- Author: Mohammad Khelghati
- Published: Jul 29, 2026
- Type: Problem & solution

ProdLine CoPilot addresses in-shift production-line decisions when equipment faults threaten output and data is split across PLCs, SCADA, MES, ERP, and LIMS. The proposed system streams operational technology data into Databricks, joins it with slower business and quality records, and routes questions to domain specialists that calculate recovery, depletion, quality-risk, and maintenance signals. Its orchestrator reads current Unity Catalog state, uses SQL, Genie, AI Search, Model Serving, and Python-based tools, while MILP, stochastic scheduling, Monte Carlo, Bayesian risk, and Pareto methods provide analytical outputs. Recommendations are tested against 1,000 scheduling scenarios and returned with trade-offs in cost, overtime, throughput, quality, and service, but execution remains with line managers, quality leads, and maintenance. The demo drafts work orders, quality deviations, and schedule updates for approval, with traceability storing inputs, assumptions, constraints, approvers, and outcomes.


### [How agentic AI can help telecom finance teams protect the margin when every moment matters](https://yomu.fyi/post/how-agentic-ai-can-help-telecom-finance-teams-protect-the-margin-when.md)
- Company: [Databricks](https://yomu.fyi/company/databricks.md)
- Author: Elena Tesser
- Published: Jul 29, 2026
- Type: Explainer

Telecom finance teams face revenue leakage when services go unbilled or uncollected, fraud and partner-settlement errors drain charges, and customer churn removes future spending. The source argues that fragmented ordering, billing, ERP, and spreadsheet data makes month-end reconciliation too slow, turning discrepancies found weeks later into write-offs rather than recoverable revenue. It presents Databricks Genie as a data-smart AI coworker whose ontology captures business meaning across systems, stays current, and grounds sourced answers in governed data. Genie can answer billing, invoice, spend, and variance questions, surface anomalies, and support action while a person remains responsible for decisions; Lumen Technologies is cited as using it across the Office of the CFO. The text says Genie does not make billing, fraud, or pricing decisions.


### [How to Self-Host a Validated AI Coding Assistant with NVIDIA NeMo Guardrails](https://yomu.fyi/post/how-to-self-host-a-validated-ai-coding-assistant-with-nvidia-nemo-guar.md)
- Company: [NVIDIA](https://yomu.fyi/company/nvidia-developer-blog.md)
- Author: Tanya Lenz
- Published: Jul 29, 2026
- Type: Tutorial

Deploying AI coding assistants in regulated or sovereign environments presents risks around data privacy, hallucinated dependencies, and auditability. This tutorial outlines a self-hosted architecture deploying StarCoder2-7B as an NVIDIA NIM endpoint served entirely from local GPUs. NVIDIA NeMo Guardrails sits as an OpenAI-compatible proxy between developer IDEs and the model to intercept and reject requests for restricted or human-only code paths. Downstream CI gates catch package hallucinations and verify changes, while commit trailers and a Prometheus and Grafana metrics loop monitor defect escape rates. Because the model remains distinct from the external policy and validation layers, teams can incrementally adopt components or upgrade models without rearchitecting the pipeline.


### [Tame Dependabot: Group your updates, slow the cadence, keep security fast](https://yomu.fyi/post/tame-dependabot-group-your-updates-slow-the-cadence-keep-security-fast.md)
- Company: [Github](https://yomu.fyi/company/github.md)
- Author: Bruno Borges
- Published: Jul 29, 2026
- Type: Problem & solution

Dependabot often floods repository maintainers with individual daily pull requests for single patch bumps, wasting review and continuous integration resources. Microsoft's GCToolkit project mitigated this issue by updating its dependabot.yml configuration to group dependency updates using wildcard patterns and slowing the schedule interval to monthly. The project also expanded coverage to include Maven alongside GitHub Actions so all relevant package ecosystems receive managed updates. Because Dependabot processes security alerts independently of regular version schedules and enforces a default three-day package cooldown, critical vulnerability fixes remain fast while routine maintenance noise drops significantly.


### [A new benchmark for evaluating patient-facing health AI agents](https://yomu.fyi/post/a-new-benchmark-for-evaluating-patient-facing-health-ai-agents.md)
- Company: amazon.science
- Author: Korosh Vatanparvar, Ashutosh Joshi
- Published: Jul 29, 2026
- Type: Benchmark

Healthcare AI benchmarks often rely on static medical knowledge questions or clinician-facing tool evaluations, failing to test multiturn patient interactions and clinical workflow execution. To address this gap, PatientAgentBench introduces an automated evaluation pipeline using synthetic patient charts, dual-agent conversations, and an LLM-as-a-jury panel. The evaluation framework applies over one hundred clinician-vetted criteria across six dimensions, including clinical safety, triage quality, and workflow accuracy. Testing frontier models revealed a severity paradox where agents scored higher on obvious emergencies than on routine administrative requests harboring hidden risks. Common safety failures included omitting crisis hotline resources and fabricating clinical information, showing that raw foundation model capability alone cannot guarantee safe patient-facing execution.


### [Developers are attached to tools because tools encode trust](https://yomu.fyi/post/developers-are-attached-to-tools-because-tools-encode-trust.md)
- Company: [Stack Overflow](https://yomu.fyi/company/stack-overflow.md)
- Author: Ryan Donovan
- Published: Jul 29, 2026
- Type: Explainer

Traditional developer tools like Vim, Emacs, and IDEs earn trust through predictability, bounded roles, and muscle memory developed over time. In contrast, agentic coding tools operate across the entire software development lifecycle using ambiguous natural language, rapidly generating vast amounts of code that developers do not inherently trust. This influx shifts development bottlenecks from code authoring to code review, validation, and infrastructure cost management. Tooling additions such as automated code reviewers, context managers, and memory systems cannot repair broken software development lifecycles unless organizations actively adapt their culture and workflows. Teams can establish trust in agentic environments by providing explicit contextual specifications, reusing verified components to avoid duplicate implementations, and restricting non-deterministic artificial intelligence to appropriate problem domains.


### [Security update (July ‘26)](https://yomu.fyi/post/security-update-july-26.md)
- Company: [Notion](https://yomu.fyi/company/notion.md)
- Author: Daniel Pyykonen
- Published: Jul 29, 2026
- Type: Incident / postmortem

Notion investigated public claims that a large dataset of its data was offered for sale on X and Breached Forums. The investigation began after the claim surfaced on June 23, 2026, and included review of publicly shared materials, sample data, and support from a third-party partner. The sample did not match data in Notion systems and contained internal duplicates and inconsistent fields; the claimed asking price and the forum user’s history also raised concerns. A deeper investigation found further inconsistencies indicating that the alleged dataset could not have come from Notion’s systems, and forum administrators later flagged the user as fraudulent. Based on the reviewed information, Notion found no evidence of a breach, customer-data access, or compromise, concluding that the alleged materials were fabricated.


### [Sign in with ChatGPT is in beta on Supabase](https://yomu.fyi/post/sign-in-with-chatgpt-is-in-beta-on-supabase.md)
- Company: [Supabase](https://yomu.fyi/company/supabase.md)
- Author: cemal\_kilic, gregor\_vand
- Published: Jul 29, 2026
- Type: Announcement

Supabase is rolling out Sign in with ChatGPT in beta on its login page and through the Supabase plugin in ChatGPT on desktop, web, and mobile. The flow lets new users create a Supabase account with their ChatGPT identity, while existing Supabase accounts with the same email are linked automatically; SSO accounts remain separate. Users can add Supabase from ChatGPT’s plugins directory, authenticate with ChatGPT, review a consent screen, and return to the conversation with the connection enabled. Authentication and plugin authorization remain separate, with access revocable from the Supabase Dashboard, and users must create an organization in the Dashboard before creating projects from ChatGPT. The beta is described as a partnership with OpenAI intended to reduce account setup between ChatGPT or Codex and a Supabase backend.


### [Cursor, now on iPad](https://yomu.fyi/post/cursor-now-on-ipad.md)
- Company: [Cursor](https://yomu.fyi/company/cursor.md)
- Published: Jul 29, 2026
- Type: Announcement

Cursor is now available on iPad for all paid plans, extending its mobile workflow for creating, reviewing, and merging pull requests. The iPad layout uses the larger display with pinned sidebar chats, split-screen review alongside chat, full-file diffs, and expanded markup that supports screenshot comments and Apple Pencil drawing. On both iPhone and iPad, the review screen covers the full PR, including comments, checks, approvals, reviewer changes, and prompts to an agent to resolve comments. An Inbox organizes active work, attention items, and PRs under review, while additional updates add Bitbucket and Azure DevOps SCM support, multi-PR sessions, and in-app switching between teams.


### [Together AI announces strategic partnership with Moonshot AI to natively serve Kimi models](https://yomu.fyi/post/together-ai-announces-strategic-partnership-with-moonshot-ai-to-native.md)
- Company: [Together AI](https://yomu.fyi/company/together-ai.md)
- Published: Jul 29, 2026
- Type: Announcement

Together AI announces a strategic partnership with Moonshot AI to serve Kimi models as a launch platform, beginning with Kimi K3 and extending to future open-weights releases. Kimi K3 is described as a 2.8-trillion-parameter sparse Mixture-of-Experts model with native vision and a 1M-token context window, using Kimi Delta Attention, Attention Residuals, per-head Muon, and quantile-based expert load balancing. Together says developers can access the models through Serverless, Provisioned Throughput, and Dedicated Inference, with US-hosted infrastructure, zero data retention, autoscaling, and a 99% uptime SLA for Provisioned Throughput. The partnership also supports post-training through full-weight and LoRA reinforcement learning and supervised fine-tuning, with checkpoints deployable directly to inference; Kimi K3 is available at launch, while future Moonshot releases are planned for the same platform.


### [Configuring Dedicated Model Inference](https://yomu.fyi/post/configuring-dedicated-model-inference.md)
- Company: [Together AI](https://yomu.fyi/company/together-ai.md)
- Published: Jul 29, 2026
- Type: Tutorial

Dedicated Model Inference is organized around three parts: an endpoint, deployments that run a model revision with replicas, and immutable configs that define how it runs. A deployment binds one model revision to one config and an autoscaling policy, while the endpoint provides a stable API-facing name and traffic split. Routing uses each deployment’s weight multiplied by its ready-replica count, so equal weights produce a 25%/75% split when deployments have one and three ready replicas; unavailable replicas contribute no capacity. The configuration workflow uses certified model-and-config profiles, selectable by accelerator, GPU count, optimization target, and topology, while new config revisions preserve tested rollback targets. This capacity-aware design makes scaling, rollouts, A/B tests, shadow traffic, and zero-downtime changes possible, but raw API or SDK deployments must be added to the traffic split before they receive requests.


[Newer posts](https://yomu.fyi/posts/page/85.md) · [Older posts](https://yomu.fyi/posts/page/87.md)