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
Our position on open-weights models
Anthropic CEO Dario Amodei states that Anthropic has never advocated banning open-weights models, while distinguishing models without dangerous capabilities as a public good from risks associated with sufficiently capable systems. The post identifies two concerns: authoritarian governments gaining military superiority or enabling repression, and powerful models being misused for cyberattacks, biological attacks, or serious alignment problems. It argues open weights can increase misuse risk because guardrails and monitoring are difficult after release, but says banning their use by US businesses would not address bad actors. Instead, Amodei supports restricting powerful chips and chipmaking equipment exports to China, disrupting industrial-scale distillation, and requiring global pre-release safety testing for sufficiently capable open and closed models. He agrees open weights can expand access, competition, and customer control, but says claims about safeguards should be tested empirically.
Coding Agent Horror Stories: The 29 Million Secret Problem
Compromised versions of the Nx npm package, dubbed the s1ngularity campaign, used a post-install hook to repurpose installed AI coding agents as credential scanners. Because AI command-line interfaces often run with the developer's full user permissions, the malware invoked tools such as Claude Code, Gemini CLI, and Amazon Q with permission-bypass flags to map sensitive files. GitGuardian findings indicate that AI-assisted code leaks secrets at roughly twice the baseline rate, largely because models ingest live credentials into working contexts. Docker Sandboxes addresses this risk by restricting the execution environment to the project workspace and keeping actual credentials on the host. When sandboxed with sbx, agents receive sentinel placeholders, while an egress proxy injects real secrets into outgoing network requests.
How Vercel used Cursor to build Queues
Vercel built its durable event streaming system, Vercel Queues, entirely with Cursor, using agents across three rewrites and validating changes against real AWS environments. Engineers had agents work on the Queues API, connect to AWS, inspect DynamoDB and Kinesis side effects, inject faults, simulate bugs, and verify results, while Cursor retained context through long sessions and compaction events. The workflow also supported model switching, parallel planning agents, inline review, stacked PRs, and asynchronous Cloud Agents. Among Vercel’s Cursor power users, PR throughput increased 54%, PR cycle time decreased 89%, resources per feature fell 27%, and the company reports 104 hours saved per developer annually. Product managers also use Cursor for repository analysis, product analytics, and small fixes without drawing on engineering resources.
Introducing Cursor Start
Cursor has launched Cursor Start, a monthly plan for developers in India priced at ₹649, with tax included and payment through UPI or card. The plan responds to requests for India-specific pricing and UPI support, and follows a reported tripling of Cursor’s Indian user base to more than 3M developers, making India its third-largest market globally. Start provides generous access to Grok 4.5 and Composer, more agent requests than the Free plan across desktop, web, iOS, and CLI, plus always-on cloud agents, Cursor for iOS, and workflow extensions through plugins, MCP servers, hooks, and skills. It sits between Free and Pro, offering everyday building capabilities while Pro adds broader model access and features such as Bugbot, Auto mode, Automations, and the Cursor SDK.
Build a Flask App with Okta for Secure OIDC Login and Authorized API Calls
Flask web applications require dedicated mechanisms to authenticate users securely and authorize access to downstream services. Developers can address this by integrating Okta as an OpenID Connect identity provider using the Authlib library. Authlib registers Okta as a remote provider and automates the Authorization Code flow using Proof Key for Code Exchange with SHA-256 code challenge verification. After logging in through Okta and completing the callback route, the Flask web application uses the acquired OAuth 2.0 access token to query a protected backend resource server. This architecture ensures that user credentials remain managed by Okta while client requests to backend API endpoints carry valid, scoped authorization tokens.
Get Started with Genie One: Top AI Cowork Use Cases for Business Users
Genie One is presented as a data-smart, agentic coworker for business users who need work completed across calendars, CRMs, data warehouses, ticketing systems, and shared documents rather than through a standalone chatbot. The guide explains four workflows: recurring business reviews, meeting preparation and follow-up, knowledge-work document automation, and operational monitoring with alerts. Users connect relevant sources, provide templates or natural-language rules, create reusable skills, test outputs, schedule routines, and refine instructions over time; the monitoring workflow can compare metrics with defined thresholds and email alerts. The post recommends starting with one narrow recurring task, connecting its required sources securely with data or IT partners, and iterating toward a library of saved skills that can be used across the organization.
The EU Digital Product Passport: a traceability deadline
The EU’s Ecodesign for Sustainable Products Regulation (ESPR) will require in-scope products sold in the single market to carry a machine-readable Digital Product Passport (DPP), with batteries first from February 2027. The passport links a unique identifier and data carrier to lifecycle information that must remain accurate, while operators retain responsibility for data stored on their own backend, tiered access, supplier inputs, and any Registry registration. Implementation is primarily a data-integration and governance problem, spanning tier-N suppliers, per-unit operational records, lineage, analytics, AI, and controlled sharing rather than QR-code generation. Databricks components including Lakebase, Unity Catalog, Lakeflow Spark Declarative Pipelines, Databricks Apps, and OpenSharing are presented with a battery-focused Solution Accelerator as an operator-side reference architecture. The stated benefits extend beyond market access to faster recalls, sourcing-risk visibility, and lineage-based sustainability reporting, but technology alone does not establish compliance.
Building Reliable Production AI with Durable Workflows
AI prototypes rely on simple prompt-response interactions, but production AI systems handling thousands of records must manage distributed execution issues like worker crashes, rate limits, and transient failures. Rather than treating long-running operations as single large jobs or rebuilding state from logs, systems require durable workflows that preserve execution history across failures. In Agentforce Grid, execution is organized into a hierarchy where column runs divide into recoverable batches and rows composed of discrete activities. Each batch checkpoints after every successful row, retrying up to ten times with exponential backoff and resuming strictly at incomplete work to prevent duplicated model calls. Exposing progress across these hierarchical boundaries ensures that both users and engineers can distinguish completed, failing, and retryable units.
Indexing the Data Lake for Online Point Queries
Online services and AI agents require low-latency point queries over massive datasets that are economically impractical to store entirely in key-value systems like Bigtable or DynamoDB. While cloud object storage offers rapid request latency, traditional distributed query engines like Trino and BigQuery introduce seconds of planning and scheduling overhead along with dependent file reads. Random Access Parquet addresses this limitation by using an external index mapping keys directly to exact Parquet files and row locations. Readers execute O(1) index lookups and issue parallel ranged reads to fetch only the required bytes without scanning footers or metadata. By running directly against existing data lake files, this architecture eliminates separate storage copies and allows shared analytical data to serve interactive online lookups.
Pinner Progression: Better Use-Case Representation Driving Weekly Active User Growth at Pinterest
Pinterest designed the Pinner Progression initiative to shift its home feed recommendation pipeline from short-term engagement optimization toward long-term user retention. Standard ranking systems model user actions as immediate sequential features, failing to capture the evolving lifecycles of distinct user habits. To address this gap, the platform introduced User Interest Clusters (UICs), which cluster a user's engaged Pins in OmniSage embedding space using dynamic cluster counts and stateful lifecycle metadata. Pinterest integrated a UIC-aware penalty directly into Sliding Spectrum Diversity (SSD) scoring by matching Pin embeddings to cluster medoids via cosine similarity. Online experiments demonstrated that balancing use-case representation increased interaction diversity, extended session lengths, and delivered meaningful engagement gains across the platform.
NVIDIA Ising Enables Fully Automated Quantum Computer Calibration with Enhanced In-Context Learning
NVIDIA Ising Calibration is an open-source vision language model designed to interpret quantum processor diagnostic outputs and recommend tuning steps. The newly released Ising Calibration 1.5 features a 31-billion-parameter architecture that is 11.4% smaller at BF16 precision than its predecessor. Trained on multi-modality partner data including superconducting qubits, quantum dots, ions, neutral atoms, and electrons on Helium, the model supports both zero-shot reasoning and in-context learning. Evaluated on the QCalEval benchmark, it scores 10% higher on average than comparable open models in zero-shot tasks and improves in-context learning performance by over 86% relative to the prior release. In addition to BF16 weights, the model offers an NVFP4 quantization option suitable for deployment on single consumer GPUs or NVIDIA DGX Spark hardware.
Expanding our partnership with Cognizant | Anthropic
Anthropic is expanding its partnership with Cognizant, a technology services company, to deploy Claude more broadly across enterprise work. Cognizant is embedding Claude in its business and engineering platforms, scaling a Claude-certified workforce through its Frontier Certified model, and becoming a Global Premier Partner in the Claude Partner Network. Its Flowsource platform runs Claude Code alongside engineers in a Spec-Driven Development module, directing outputs with project specifications, coding standards, and architectural blueprints before evaluating them for production. The company reports more than 30,000 associates have completed Claude training and describes client deployments including a contract-intelligence system that cut review time by up to 40% and raised extraction accuracy above 88%, plus a risk-navigation tool that saves underwriters roughly eight hours a week. The collaboration combines Cognizant’s industry context, engineering scale, and delivery capabilities with Claude for production enterprise use.
AI Sovereignty is Your Alpha: How to Avoid Transferring Your Alpha to a Hosted Model Provider
The document examines how use of third-party AI model services can expose an organization’s “alpha”—institutional knowledge and tradecraft embodied in data sent to and generated from models—to hosted model providers. It identifies risks involving provider reuse, storage, human review, model or service improvement, contractual liability caps, ZDR-ineligible beta services, safety-classifier triggers, and inference routed through prohibited regions. Its recommended approach combines contractual protections with operational controls: define robust Zero Data Retention, limit data use to service delivery, bind both hyperscalers and relevant AI labs, maintain an allow-list, block beta headers, and constrain unilateral term changes. It also calls for observability and contractual incident safeguards, including investigation periods, migration buffers, and regional routing limits, while noting that the document is not legal advice.
NVIDIA Cosmos-H-Dreams: Bringing Real-Time Generative Simulation to Surgical Robotics
Surgical robotics needs safer, faster ways to train and evaluate vision-language-action policies because physical experiments are costly, slow to reproduce, and vulnerable to damaging instruments or biological material. NVIDIA introduces Cosmos-H-Dreams, a real-time, action-conditioned generative simulator that specializes Cosmos-H-Surgical-Simulator for da Vinci Research Kit tabletop suturing and generates future frames from an initial RGB image and live robot kinematics. Its causal student is trained through teacher fine-tuning, causal warmup, and self-forcing distillation, including failure episodes and long autoregressive horizons, while few-step diffusion reduces generation cost. Served through FlashDreams with streaming KV caching, CUDA Graph capture, and model compilation, it reaches approximately 160 frames per second on one NVIDIA RTX PRO 6000, compared with roughly 10 frames per second for standard inference. The system supports human or learned-policy control, but is presented as a research and development platform rather than a diagnostic system or physical-robot controller.
Six Agent Harness Capabilities for Higher Model Performance
NVIDIA Labs released NVIDIA Labs Object-Oriented Agents (NOOA), an open-source research preview treating AI agents as single Python classes. The harness implements six model-facing capabilities: typed inputs and outputs, passing live objects by reference, code-based actions, programmable loops, explicit object state, and model-callable harness APIs. Instead of context compaction, passing live references keeps transcripts append-only, preserving prompt cache validity across long sessions. NOOA also includes a curated SQLite-backed long-term memory system forming a typed knowledge graph. Across evaluations, NOOA achieved 82.2 percent on SWE-bench Verified with GPT-5.5 using roughly 1.1M tokens per task, 86.8 percent on CyberGym L1 without network access, and 85.1 percent mean RHAE on ARC-AGI-3 using GPT-5.6-sol at under twenty dollars per game.
AI customer service: strategy, agents, and solutions guide
AI customer service combines natural language processing, machine learning, predictive analytics, generative AI, and automation to route requests, generate responses, and support resolution across channels. The guide distinguishes AI agents from scripted chatbots: agents reason over context, call external tools, and complete multi-step tasks such as checking shipping data and issuing a partial refund. It presents AI as a layer that absorbs routine volume while human agents retain judgment over complex or sensitive cases, citing potential gains including more than 70% query automation, up to 30% lower operating costs, and a 15% improvement in customer satisfaction. It recommends evaluating integration, decision transparency, production support, security, and compliance, then starting with one measurable, high-volume use case and expanding only after consistent resolution quality.
Building AI Literacy: Frameworks, Tools, and Practices
AI literacy is presented as the ability to understand how AI works, use its tools effectively, evaluate outputs critically, and recognize ethical implications. The guide organizes this fluency into functional, critical, and ethical domains, with progression from defining terms and crafting prompts to comparing tools, auditing outputs for bias, and designing human oversight. It explains that large language models generate probable continuations from statistical patterns in training data, so hallucinations and bias require verification against trusted sources rather than assuming factual correctness. For education and organizations, it recommends distributed curriculum or role-specific modules, hands-on assignments, measurable assessment checkpoints, governance policies, internal champions, iterative pilots, and longer-term impact metrics.
NVIDIA Nemotron 3 Ultra Leads Open Models on Accuracy and Efficiency in Agentic RTL Coding
Register transfer level (RTL) development relies heavily on iterative verification and tool feedback, which poses challenges for standard code generation models. To address realistic hardware design tasks, the ACE-RTL agent pairs an iterative generate-test-reflect workflow with NVIDIA Nemotron 3 Ultra. The model utilizes a hybrid Mamba-Attention Mixture-of-Experts architecture trained on a synthetic dataset encompassing code generation, editing, and error debugging. Across the comprehensive Verilog design problems benchmark, ACE-RTL with Nemotron 3 Ultra attained a 97.1% average pass rate, outperforming GLM 5.2 and Kimi K2.6. Additionally, Nemotron 3 Ultra achieved its accuracy while consuming an average of 6,629 tokens per iteration, which is substantially lower than competing models.
Advancing Semiconductor Innovation Across Materials Engineering and Manufacturing
Increasing artificial intelligence workloads and slowing geometric scaling require rapid semiconductor innovation across materials engineering and manufacturing. Applied Materials collaborated with NVIDIA to build an end-to-end digital development pipeline that accelerates atomic-scale modeling, process development, and factory optimization. The Ginestra defect-centric simulation platform integrates NVIDIA cuDSS sparse solvers to achieve up to a 10x speedup, while NVIDIA cuEST accelerates density functional theory calculations on B200 systems by approximately 55x compared to 64 CPU cores. In process engineering, the ACE+ multiphysics platform pairs with NVIDIA PhysicsNeMo to run topography simulations up to 35x faster, and NVIDIA Omniverse generates fab digital twins to optimize material flows. This unified digital workflow enables engineers to perform virtual experiments rapidly, mitigate production bottlenecks, and shorten transitions to high-volume manufacturing.
Kimi K3 by Moonshot now available on Modal
Moonshot released Kimi K3, a 2.8 trillion parameter multimodal mixture-of-experts transformer with a 1 million token context window and native vision, now on Modal. Modal runs the model at 460 tokens per second on release day through a partnership with Moonshot and vLLM, with token-based pricing on a Shared API or dedicated Auto Endpoint. K3 activates 16 of 896 experts per token and uses Kimi Delta Attention along with Attention Residuals to achieve roughly 2.5 times the scaling efficiency of its predecessor K2. Moonshot employed quantization-aware training from the SFT stage with MXFP4 weights and MXFP8 activations, and contributed a new prefix caching implementation to vLLM to address challenges from KDA. Modal deployed a custom DFlash speculative decoding drafter tuned to K3’s architecture, yielding 360 percent faster interactivity and 88 percent higher throughput on agentic workloads.
Top topics
- Architecture834
- Performance816
- LLMs660
- Machine Learning622
- Open Source563
- AI Agents553
- AI470
- Developer Experience448
Weekly digest
The best engineering reads, once a week.