# cost Published articles for cost. This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles. ## Introducing Gemini 3.8 Live and 3.8 Live Extended Thinking DevFeed: [Introducing Gemini 3.8 Live and 3.8 Live Extended Thinking]() Original publisher: [Read original article]() Author: Tom Ouyang Published: 2026-09-15T17:05:57Z Content type: release Language: en Sources: [Google DeepMind News]() Topics: [Artificial Intelligence](), [real-time](), [speech-to-speech](), [Google](), [API]() Tags: [ai](), [api](), [audio](), [cost](), [efficiency](), [gemini](), [google](), [none](), [real-time](), [speech-to-speech](), [tools]() ### AI overview Google introduces Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking, two models designed for near-real-time voice interaction and reasoning. The release describes visual grounding, multilingual conversation, background tool and API execution, and deeper reasoning for complex workflows. ### Source excerpt Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking are our most advanced live dialogue models yet, built for natural conversation. ## Dense vs. MoE Models: Active Parameters, Throughput, and When to Choose Each DevFeed: [Dense vs. MoE Models: Active Parameters, Throughput, and When to Choose Each]() Original publisher: [Read original article]() Author: Elizabeth Goodman Published: 2026-09-15T17:00:11Z Content type: article Language: en Sources: [NVIDIA Developer](), [NVIDIA Technical Blog]() Topics: [Mixture of Experts (MoE)](), [Nemotron](), [Routing (disambiguation)]() Tags: [agentic-ai-generative-ai](), [ai-inference](), [architecture](), [cost](), [data-center-cloud](), [deployment](), [llms](), [memory](), [mixture-of-experts-moe](), [models](), [moe](), [nemo](), [nemotron](), [performance](), [router](), [routing](), [tokens]() ### AI overview This article explains how dense and Mixture-of-Experts models activate parameters, compares their effects on throughput, memory cost, and serving complexity, and discusses when each architecture fits different deployment constraints. It uses Nemotron 3.5 Lightning as an example of an MoE model. ### Source excerpt How can a 30B-parameter model activate only 3B parameters per token, and still use the capacity of the larger model? Nemotron 3.5 Lightning illustrates the... ## Build an AI-powered product tagging system with Amazon SageMaker serverless model customization DevFeed: [Build an AI-powered product tagging system with Amazon SageMaker serverless model customization]() Original publisher: [Read original article]() Author: Linpo Guo Published: 2026-09-15T16:11:36Z Content type: tutorial Language: en Sources: [Artificial Intelligence]() Topics: [Artificial Intelligence](), [Amazon SageMaker AI](), [Fine-tuning](), [rlvr](), [grpo](), [Inference](), [model-deployment](), [Reinforcement learning](), [SDK](), [Serverless]() Tags: [ai](), [amazon-sagemaker](), [amazon-sagemaker-ai](), [aws](), [compute](), [cost](), [customization](), [expert-400](), [fine-tuning](), [inference](), [jobs](), [reinforcement-learning](), [rlvr](), [sdk](), [technical-how-to]() ### AI overview This walkthrough shows how to build a product tagging system by customizing Qwen3-8B with supervised fine-tuning and reinforcement learning with verifiable rewards on Amazon SageMaker serverless model customization. It then deploys the optimized model for asynchronous inference to enrich retail catalogs. ### Source excerpt Manually tagging thousands of catalog products is slow and inconsistent. This walkthrough shows how to customize Qwen3-8B with supervised fine-tuning (SFT) and reinforcement learning with verifiable rewards (RLVR) on Amazon SageMaker serverless model customization, then deploy it for asynchronous inference to build a cost-efficient product tagging system. ## How LLMs Handle Memory Through Context and Surrounding Applications DevFeed: [How LLMs Handle Memory Through Context and Surrounding Applications]() Original publisher: [Read original article]() Author: ByteByteGo Published: 2026-09-15T15:31:12Z Content type: article Language: en Sources: [ByteByteGo]() Topics: [Large Language Model](), [context window](), [App](), [long-context]() Tags: [article](), [context](), [context-window](), [cost](), [latency](), [llms](), [memory]() ### AI overview LLMs do not usually retain personal or persistent memory between interactions. Surrounding applications create the appearance of memory by storing messages, maintaining summaries, retrieving relevant information, and supplying it to the model. As conversations grow, this processing increases cost and latency, while context-window limits require older information to be removed, summarized, or stored elsewhere. ### Source excerpt In this article, we will learn how LLMs handle memory so that they are useful to end users in performing complex tasks that require conversation and holding context. ## Syitren R400 is a portable Bluetooth CD player with a modular, customizable design DevFeed: [Syitren R400 is a portable Bluetooth CD player with a modular, customizable design]() Original publisher: [Read original article]() Author: Ida Torres Published: 2026-09-15T13:20:31Z Content type: opinion Language: en Sources: [Yanko Design]() Topics: [Hardware](), [Bluetooth]() Tags: [2026](), [a-design-award-and-competition](), [audio](), [audio-music-technology-a-design-award-and-competition-cd-player-music-player](), [bluetooth](), [cd-player](), [collection](), [cost](), [design](), [hardware](), [modular](), [mount](), [music](), [music-player](), [portable](), [product](), [technology]() ### AI overview The article presents Syitren's R400 as an entry-level portable Bluetooth CD player designed by ORGDOT Design. Its detachable modular shell supports customization, a desk stand, and an optional wall-mount bracket, while dual-material construction and cable-management details aim to improve everyday use. ### Source excerpt Finally, a CD Player Gen Z Actually Wants to Own It really warms my heart when I see Gen Z and Gen Alpha kids collecting physical media from their favorite artists. These are people who... ## LF Energy Research Finds Open Source Software Can Deliver 2-5x Greater Net Value for Grid Operators DevFeed: [LF Energy Research Finds Open Source Software Can Deliver 2-5x Greater Net Value for Grid Operators]() Original publisher: [Read original article]() Author: andrewb@proximabiz.com (The Linux Foundation) Published: 2026-09-15T07:00:00Z Content type: news Language: en Sources: [Linux Foundation - Blog]() Topics: [Open Source](), [Software](), [cloud-infrastructure](), [digital sovereignty](), [interoperability](), [Simulation]() Tags: [ai-data-centers](), [case-studies](), [compare](), [cost](), [digital-sovereignty](), [framework](), [infrastructure](), [interoperability](), [linux-foundation](), [open-source](), [renewable-energy](), [report](), [research](), [september-2026](), [software](), [summit]() ### AI overview LF Energy reports that open source software can provide grid operators with 2-5 times greater net value than conventional software procurement. Its Open Source Benefit-Cost Framework evaluates total cost of ownership, risk exposure, strategic value, and societal impact using case studies and simulations. ### Source excerpt New benefit-cost framework gives utilities and regulators a standardized methodology to compare open source with conventional software procurement ## Gemini 3.8 Live models now available on AI Gateway DevFeed: [Gemini 3.8 Live models now available on AI Gateway]() Original publisher: [Read original article]() Author: Jerilyn Zheng Published: 2026-09-15T00:00:00Z Content type: release Language: en Sources: [Vercel News]() Topics: [real-time](), [Artificial Intelligence](), [API](), [Google](), [WebSocket](), [SDKs]() Tags: [ai](), [api](), [cost](), [failover](), [gemini](), [google](), [real-time](), [realtime](), [sdk](), [websocket]() ### AI overview Vercel announces that Google's Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking models are available on AI Gateway. The models support real-time spoken interactions, audio and visual grounding, multilingual switching, background tool calls, and parallel reasoning through the AI SDK's realtime API. ### Source excerpt Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking from Google are now available on AI Gateway. Both models support real-time spoken interactions for voice assistants, conversational experiences, and applications that respond through audio. google/gemini-3.8-live supports real-time audio, visual grounding, automatic switching across 97 languages, and background tool calls while the conversation continues. google/gemini-3.8-live-extended-thinking adds multi-step reasoning that runs in parallel with speech, allowing it to acknowledge requests and narrate progress without interrupting the conversation. Use either model through the AI SDK's realtime API. Install the Gateway provider and a WebSocket client: Mint a short-lived token, open the WebSocket, and use the model adapter to serialize and parse realtime events: See the realtime quickstart for more details on realtime events and WebSocket connections. Try Gemini 3.8 Live or Gemini 3.8 Live Extended Thinking in the model playground. AI Gateway provides a unified API for calling models, tracking usage and cost, and configuring retries, failover, and performance optimizations for higher-than-provider uptime. Read more ## Apple Laid Off Around 100 Staff Working On Vision Pro, But Don't Panic DevFeed: [Apple Laid Off Around 100 Staff Working On Vision Pro, But Don't Panic]() Original publisher: [Read original article]() Author: David Heaney Published: 2026-09-14T13:06:59Z Content type: news Language: en Sources: [UploadVR]() Topics: [visionOS](), [Security](), [Meta]() Tags: [apple](), [article](), [company](), [cost](), [ecosystem](), [gaming](), [immersive-video](), [industry-news](), [meta](), [open](), [report](), [security](), [series](), [third-party](), [video](), [vision]() ### AI overview Apple laid off around 100 Vision Pro staff, largely shutting down its spatial gaming team and reducing its immersive video team. The cuts also affected security, spatial audio, Siri integration, and visionOS personnel, but Apple is not leaving the VR/MR market. The changes reflect limited gaming use and the high cost of producing immersive video. Apple may launch a slimmer, lighter successor headset around late 2028, although its continuation could depend on the wider headset market. ### Source excerpt Apple laid off around 100 staff working on Vision Pro, including "largely shutting down" the spatial gaming team and downsizing the immersive video team, Bloomberg's Mark Gurman reported. ## Which Video Streaming Provider Is Best for Corporate Training? DevFeed: [Which Video Streaming Provider Is Best for Corporate Training?]() Original publisher: [Read original article]() Author: Max Wilbert Published: 2026-09-14T12:40:20Z Content type: comparison Language: en Sources: [DaCast]() Topics: [Streaming](), [communications]() Tags: [compliance](), [corporate](), [cost](), [distributed](), [live-streaming](), [streaming](), [the-video-experts-blog](), [training]() ### AI overview This guide compares Brightcove, Kaltura, Panopto, and Dacast as video streaming providers for corporate training. It focuses on LMS compatibility, compliance recording retention, and cost at scale, and summarizes the providers' stated features and pricing. ### Source excerpt By Dacast Editorial Team | Reviewed by Jon Whitehead, COO at Dacast | Updated September 2026 Training a distributed workforce is one of the hardest logistics problems corporate L&D teams face, and live streaming video has become one of the most effective ways to solve it. Choosing the right video streaming provider for training is [...] The post Which Video Streaming Provider Is Best for Corporate Training? appeared first on Dacast. ## Using Exact-Match Response Caching to Reduce LLM Costs DevFeed: [Using Exact-Match Response Caching to Reduce LLM Costs]() Original publisher: [Read original article]() Author: Abhilash Rao Mesala Published: 2026-09-14T11:00:00Z Content type: article Language: en Sources: [The New Stack]() Topics: [Caching](), [Large Language Model]() Tags: [ai-engineering](), [ai-infrastructure](), [caching](), [contributed](), [cost](), [finops](), [generation](), [hash](), [llm](), [token]() ### AI overview The article explains how to reduce LLM costs by fingerprinting requests, context, model settings, and underlying data to create exact-match cache keys. Valid cached responses can be reused without calling the model. It distinguishes response caching from provider prompt caching, where only eligible prompt computation is reused. ### Source excerpt An LLM can answer the same question a thousand times and charge you each time. Before paying for another answer, The post Why an old caching trick is your secret to lower LLM costs appeared first on The New Stack. ## Infineon RISC-V for Automotive at Hot Chips 2026 DevFeed: [Infineon RISC-V for Automotive at Hot Chips 2026]() Original publisher: [Read original article]() Author: Vic A Published: 2026-09-13T21:58:44Z Content type: article Language: en Sources: [ServeTheHome]() Topics: [RISC-V](), [Microcontroller](), [real-time](), [Low Latency](), [Architecture & Design](), [Security](), [IO](), [Inference Performance]() Tags: [2026](), [architecture](), [automotive](), [compute](), [cost](), [efficiency](), [infineon](), [latency](), [low-latency](), [microcontrollers](), [mobile](), [performance](), [posix](), [processors](), [real-time](), [risc-v](), [security]() ### AI overview Infineon presents RISC-V automotive processors for next-generation vehicle architectures. The article covers real-time control, low latency, power efficiency, security, zone controllers, central car computers, and heterogeneous workloads including DSP, AI inference, audio, and POSIX-based services. ### Source excerpt At Hot Chips 2026, Infineon presented a case for using RISC-V in various automotive processors in next-generation cars The post Infineon RISC-V for Automotive at Hot Chips 2026 appeared first on ServeTheHome. ## "Same mission, bigger stage": OpenAI hires Git AI founders to help Codex prove its ROI DevFeed: ["Same mission, bigger stage": OpenAI hires Git AI founders to help Codex prove its ROI]() Original publisher: [Read original article]() Author: Paul Sawers Published: 2026-09-12T14:46:53Z Content type: news Language: en Sources: [The New Stack]() Topics: [ai observability](), [cursor]() Tags: [ai](), [ai-agents](), [ai-coding](), [codex](), [cost](), [cursor](), [developer-tools](), [git](), [open-source](), [openai]() ### AI overview OpenAI hired Git AI's founders to help measure the performance, cost, and return on investment of Codex and other AI coding tools. ### Source excerpt OpenAI has hired the founders of Git AI, an open-source tool that tracks how much code is written by AI The post "Same mission, bigger stage": OpenAI hires Git AI founders to help Codex prove its ROI appeared first on The New Stack. ## Intel Reverses Decision To Cancel Their Open-Source Font Project DevFeed: [Intel Reverses Decision To Cancel Their Open-Source Font Project]() Original publisher: [Read original article]() Author: Michael Larabel Published: 2026-09-12T13:45:00Z Content type: news Language: en Sources: [Phoronix]() Topics: [Font](), [intel](), [Open Source](), [GitHub]() Tags: [cost](), [desktop-linux](), [developers](), [github](), [intel](), [linux-benchmarking](), [linux-hardware-benchmarks](), [linux-hardware-reviews](), [linux-how-to](), [linux-performance](), [linux-server-benchmarks](), [open-source](), [open-source-graphics](), [phoronix](), [phoronix-test-suite](), [source](), [ubuntu-benchmarks](), [ubuntu-hardware]() ### AI overview Intel reversed its decision to archive Intel One Mono, an open-source monospaced font for developers. The project had been archived after limited recent activity but was restored two days later and will remain maintained. The article also notes the discontinuation or archiving of several other Intel open-source projects. ### Source excerpt Amid Intel cutting back on open-source projects over the past two years as cost cutting measures at the company, their latest open-source project on the chopping block was the peculiar font they developed a few years back. The project was archived this week only to change course two days later... ## How to scale Alloy as a central telemetry gateway: capacity planning, load testing, and production lessons DevFeed: [How to scale Alloy as a central telemetry gateway: capacity planning, load testing, and production lessons]() Original publisher: [Read original article]() Author: Fatjon Nebiu Published: 2026-09-12T11:22:06.456390Z Content type: tutorial Language: en Sources: [Grafana Labs blog on Grafana Labs]() Topics: [Grafana]() Tags: [analytics](), [auth](), [best-practices](), [cloud](), [cost](), [data](), [deployment](), [enterprise](), [grafana](), [grafana-alloy](), [grafana-cloud](), [how-to](), [kubernetes](), [logs](), [metrics](), [monitoring](), [observability](), [opentelemetry](), [platform](), [production](), [prometheus](), [scale](), [telemetry](), [testing](), [traces]() ### AI overview A practical guide to scaling Grafana Alloy as a centralized telemetry gateway. It covers capacity planning, load testing, and production considerations for collecting metrics, logs, and traces and forwarding them to Grafana Cloud. ### Source excerpt Running Alloy as a single-instance sidecar is simple. Running it as a centralized gateway that absorbs the full telemetry stream of an enterprise platform--tens of millions of active series, terabytes of logs per day, and tens of thousands of trace spans per second--is a different challenge altogether. To get it right, you need deliberate capacity planning, honest load testing, and a monitoring setup that doesn't rely on the very thing you're testing. As part of the Professional Services team here at Grafana Labs, we've seen this firsthand working with customers. In this post, we'll walk you through the best practices we follow to help them find success, and we'll do so using real, anonymized data from a recent engagement. We'll cover how we sized and load tested a production Alloy central collector deployment on Kubernetes, what the numbers looked like under real stress, and how the cluster behaves today handling the full production telemetry workload for a large enterprise platform. By the end, you should have a better sense for how you can create your own central gateway for collecting telemetry in Grafana Cloud. Why a central gateway? Before diving into numbers, it's worth explaining the pattern. In a central gateway setup, all telemetry from application teams--metrics, logs, and traces--flows to a shared Alloy fleet via OTLP or native Prometheus/Loki write protocols. Alloy buffers, processes, batches, and forwards everything to Grafana Cloud. This gives you several things that per-team sidecar deployments struggle to provide: A single control plane: Auth, rate limiting, and routing in one place so application teams don't need to manage Grafana Cloud credentials Centralized buffering: Ensure a transient Grafana Cloud slowdown doesn't immediately cause data loss at the source Cost visibility: Configure the gateway to only accept telemetry data containing the label or attribute that is mandatory for cost-attribution Protocol normalization: Send OTLP, Prometheus Remote ## 4 engineering patterns behind the strongest AI Agents Challenge submissions DevFeed: [4 engineering patterns behind the strongest AI Agents Challenge submissions]() Original publisher: [Read original article]() Author: Sergio Villani Published: 2026-09-12T11:04:33.891311Z Content type: article Language: en Sources: [Google Developers Blog]() Topics: [AI Bots](), [MSP MCP]() Tags: [agents](), [ai-agents](), [cost](), [engineering](), [event](), [inference](), [latency](), [mcp](), [mcp-server](), [routing](), [validation]() ### AI overview An article on engineering patterns found in leading AI Agents Challenge submissions. It highlights bidirectional MCP, asynchronous event buses, unified validation for model fallbacks, and tiered routing to make agentic workflows more resilient, faster, and less costly. ### Source excerpt The recent Google for Startups AI Agents Challenge revealed that the most successful multi-agent systems rely on foundational software engineering patterns rather than just raw model power. Winning architectures consistently implemented bidirectional MCP for seamless inter-agent communication, async event buses for parallel execution, strict unified validation for model fallbacks, and tiered routing to minimize expensive inference calls. By prioritizing these structural practices over simple linear prompt chains, developers can build more resilient, low-latency, and cost-effective agentic workflows. ## The 6 best Sitecore alternatives from full DXP to composable headless CMS DevFeed: [The 6 best Sitecore alternatives from full DXP to composable headless CMS]() Original publisher: [Read original article]() Author: Adam Lehman Published: 2026-09-12T00:00:00Z Content type: comparison Language: en Sources: [Webflow Blog]() Topics: [Content Management System]() Tags: [complexity](), [content](), [cost](), [strategy](), [webflow]() ### AI overview A comparison of six Sitecore alternatives across full digital experience platforms, composable headless CMS options, and Webflow as a leaner visual platform. It weighs scope, technical investment, pricing, cost, complexity, and delivery speed. ### Source excerpt Compare the 6 best Sitecore alternatives for enterprise teams, from composable headless CMS to full DXP. ## 7 Best Lago Alternatives for Billing in 2026 (Open Source and Hosted) DevFeed: [7 Best Lago Alternatives for Billing in 2026 (Open Source and Hosted)]() Original publisher: [Read original article]() Author: Deepak Jangir Published: 2026-09-12T00:00:00Z Content type: comparison Language: en Sources: [Dodo Payments Blog]() Topics: [Software as a service]() Tags: [alternatives](), [billing](), [comparison](), [cost](), [incident-management](), [merchant-of-record](), [open-source](), [payment-processing](), [saas](), [self-hosting]() ### AI overview A comparison of seven Lago billing alternatives across open-source and hosted models. It focuses on predictable costs, the operational burden of self-hosting, metering depth, add-on capabilities, Merchant of Record coverage, and payment-related gaps. ### Source excerpt Compare 7 Lago billing alternatives across open-source and hosted options, weighing self-hosting cost, metering depth, and Merchant of Record coverage for SaaS and AI teams. ## OpenRouter provider fallbacks can cause inconsistent model behavior DevFeed: [OpenRouter provider fallbacks can cause inconsistent model behavior]() Original publisher: [Read original article]() Author: Simon Willison Published: 2026-09-11T22:49:18Z Content type: opinion Language: en Sources: [Simon Willison's Weblog]() Topics: [API](), [Large Language Model](), [Software]() Tags: [ai](), [ai-2-236](), [api](), [cost](), [generative-ai](), [generative-ai-1-982](), [llms](), [llms-1-948](), [model](), [models](), [news](), [openrouter](), [openrouter-32](), [providers](), [reasoning](), [serve]() ### AI overview OpenRouter's automatic provider fallbacks can produce inconsistent behavior for the same model because providers use different serving software, optimizations, and settings. The article notes differences in vision support and reasoning-effort handling, and points to provider.only and /endpoints for controlling or inspecting routing. ### Source excerpt So you want to use OpenRouter? One of OpenRouter's selling points is that it "handles fallbacks automatically and picks the most cost-effective option for each request", so you can call a single API endpoint for a model and get routed to the best available backend provider. Mohamed Moustafa points out a whole set of ways that this can cause you problems. Different providers run different serving software with different optimizations and settings, which means that the same OpenRouter endpoint can serve model requests that behave in different ways. Some providers even lack vision capability for vision models, and the way the reasoning effort option is processed can differ as well. Thankfully you can control which provider is routed to using the provider.only option. The /endpoints method returns the list of available providers for a specific model ID. Via Hacker News Tags: ai, generative-ai, llms, openrouter ## Beyond the price per token: Choosing the right OpenAI model on Amazon Bedrock for your workload DevFeed: [Beyond the price per token: Choosing the right OpenAI model on Amazon Bedrock for your workload]() Original publisher: [Read original article]() Author: Nick McCarthy Published: 2026-09-11T18:24:38Z Content type: article Language: en Sources: [Artificial Intelligence]() Topics: [AI Chat](), [AI Bots](), [ai observability]() Tags: [advanced-300](), [agent](), [agentic](), [amazon-bedrock](), [api](), [benchmarking](), [cost](), [evaluation](), [generative-ai](), [open-source](), [openai](), [technical-how-to]() ### AI overview The article presents an open-source benchmark for comparing OpenAI models on Amazon Bedrock with OpenAI API baselines by cost per correct answer, multi-turn agent trajectory cost, and deliverable quality. ### Source excerpt Comparing models on dollars per million tokens misses what production workloads actually pay for: outcomes. This post shares an open-source benchmarking harness that measures cost per correct answer, agent trajectory cost, and rubric-graded deliverable quality across OpenAI models on Amazon Bedrock. ## Evolving Pinterest's Embedding Retrieval Platform DevFeed: [Evolving Pinterest's Embedding Retrieval Platform]() Original publisher: [Read original article]() Author: Pinterest Engineering Published: 2026-09-11T15:01:03Z Content type: article Language: en Sources: [Pinterest Engineering Blog - Medium]() Topics: [Architecture & Design](), [Algorithms](), [Scalability](), [Latency](), [IO]() Tags: [ann](), [cost](), [embedding](), [embeddings](), [engineering](), [infrastructure](), [latency](), [memory](), [models](), [pinterest](), [platform](), [quantization](), [real-time](), [retrieval](), [scalability](), [search]() ### AI overview Pinterest describes evolving its Manas embedding-retrieval platform to address the cost, scale, and flexibility challenges of serving billions of embeddings. The excerpt covers ANN search, vector quantization, and SSD-based serving. ### Source excerpt Authors: Bowen Zhou | Staff Software Engineer; Shan Gao | Senior Software Engineer; Jingwen Hu | Software Engineer II; Wenjiang Chu | Staff Software Engineer The Billion-Embedding Challenge At Pinterest, the "signal" is our lifeblood. Whether it's a home decor enthusiast finding the perfect rug or a fashion seeker discovering a new aesthetic, our discovery engine relies on understanding deep semantic relationships to help our users find inspirations. Over the last few years, the explosive growth of embedding-based retrieval has fundamentally transformed how we surface these signals -- and at the heart of that transformation is Manas, Pinterest's in-house distributed search platform. Embedding Retrieval is one of the core capabilities of Manas, supporting multiple approximate nearest neighbor search algorithms, hybrid queries with both token and embedding clauses, as well as real-time updates to ensure fresh contents become searchable within seconds. Deployed on over 80 clusters and serving billions of embeddings, Manas embedding retrieval powers all major product surfaces at Pinterest including Home Feed, Search, Related Pins, Ads, and Notifications. However, as our corpus scales toward tens of billions of embeddings and our models capture increasingly complex interactions, we face mounting challenges around cost efficiency, scalability, and flexibility. On the infrastructure side, traditional ANN algorithms like HNSW are notoriously memory-hungry -- they require the entire index to reside in RAM to maintain low query latency, making cost grow linearly with corpus size. On the modeling side, the classic two-tower retrieval paradigm is too restrictive: it reduces each candidate to a single embedding and scores relevance through a simple dot product, leaving little room to express richer, context-dependent notions of similarity. To tackle these challenges, our team has been evolving Manas's embedding retrieval stack across three fronts: Quantization. We reduce the memor ## Session Traces and Cost Controls Help Diagnose AI Agent Failures DevFeed: [Session Traces and Cost Controls Help Diagnose AI Agent Failures]() Original publisher: [Read original article]() Author: Mark Silvester Published: 2026-09-11T08:14:00Z Content type: news Language: en Sources: [InfoQ]() Topics: [ai observability](), [AI Bots](), [debugging]() Tags: [agents](), [ai](), [ai-agent](), [ai-ml-data-engineering](), [artificial-intelligence](), [cost](), [debugging](), [devops](), [llm](), [loops](), [ml-data-engineering](), [monitoring](), [news](), [observability](), [observability-ai-agents](), [opentelemetry](), [traces]() ### AI overview The article describes using nested session traces, execution metrics, and cost limits to investigate and contain AI agent failures such as repeated tool calls and runaway spending. ### Source excerpt Session traces and cost controls are emerging as key observability techniques for diagnosing AI agent failures, helping teams spot tool-call loops and runaway spend while preserving enough execution context for post-incident debugging. By Mark Silvester ## DeepSeek's new model sets a template for powerful LLMs that run lean DevFeed: [DeepSeek's new model sets a template for powerful LLMs that run lean]() Original publisher: [Read original article]() Author: Tobias Mann Published: 2026-09-11T07:15:00Z Content type: news Language: en Sources: [www.theregister.com - Articles]() Topics: [deepseek](), [Large Language Model](), [Caching](), [Latency](), [Google]() Tags: [ai-and-ml](), [cache](), [compute](), [cost](), [datacenter](), [deepseek](), [embedding](), [flash](), [google](), [gpu](), [latency](), [llm](), [llms](), [model](), [release]() ### AI overview DeepSeek V4.1 Flash is a larger LLM whose architectural changes aim to reduce serving memory and compute needs. The article highlights lower KV-cache consumption, improved prompt processing, and an N-gram-based conditional memory module. ### Source excerpt DeepSeek V4.1 Flash proves that just because you build a bigger model doesn't mean you need more GPUs to serve it ## How to calculate DevOps platform total cost of ownership DevFeed: [How to calculate DevOps platform total cost of ownership]() Original publisher: [Read original article]() Author: GitLab Published: 2026-09-11T00:00:00Z Content type: tutorial Language: en Sources: [GitLab]() Topics: [CI/CD]() Tags: [ai](), [ci-cd](), [complexity](), [compute](), [cost](), [devops](), [devsecops](), [devsecops-platform](), [drivers](), [how-to](), [infrastructure](), [integration](), [model](), [platform](), [reduce](), [software-delivery](), [tools]() ### AI overview A guide to modeling the total cost of ownership of a DevOps platform, including subscriptions, CI/CD compute, AI usage, infrastructure, tools, and internal labor. ### Source excerpt There's nothing like budget pressure to put your DevOps platform under a microscope. But subscription fees and license costs only tell one part of the story. The total cost of ownership (TCO) for a DevOps platform also includes variable costs like CI/CD compute and AI usage, along with the infrastructure, tools, and employee time required to keep software delivery moving. That wider view matters when you're tasked with defending platform spend or comparing options with a head of finance. Designing a useful TCO model can: Make those costs transparent for stakeholders Shine a light on the reasoning (or lack thereof) behind each cost Identify areas to reduce spend without negatively impacting software delivery What total cost of ownership really includes The core challenge of calculating TCO is that DevOps platforms package and price capabilities differently. For example, one platform may bundle CI/CD or AI capabilities into a per-seat subscription, while another could price usage separately. A third may appear less expensive upfront but require additional tools and ongoing integration work. That's why list prices or pricing tiers alone won't give you a useful comparison. Start with the capabilities and workloads your organization actually needs, then calculate what it takes to support them on each platform. Use the same scope and time period for every option -- often one year -- and define which teams, applications, environments, and delivery stages are included. Separate recurring costs from one-time expenses and external spend from internal labor, so finance can audit the assumptions and forecast future years. A useful TCO model, therefore, answers two questions: What does it cost to meet our requirements today? Which variables will cause that cost to rise or fall as our usage changes? The cost categories that drive your bill Most DevOps platform costs fit into the following categories: Cost categoryWhat it includesMain cost driverPlatform accessPaid seats, role-based ## ToolGrad: Efficient tool-use dataset generation with textual "gradients" DevFeed: [ToolGrad: Efficient tool-use dataset generation with textual "gradients"]() Original publisher: [Read original article]() Published: 2026-09-10T22:50:22Z Content type: article Language: en Sources: [The latest research from Google]() Topics: [dataset](), [Large Language Model](), [LLM Techniques](), [Machine learning](), [Fine-tuning](), [LLM evaluation / benchmarking]() Tags: [agent](), [cost](), [datasets](), [evaluation](), [fine-tuning](), [generation](), [llm](), [machine-intelligence](), [machine-learning](), [natural-language-processing](), [prompt-engineering](), [tool]() ### AI overview ToolGrad generates tool-use chains before deriving corresponding user queries, aiming to create complex training data for LLM tool use more efficiently and at lower cost than exploration-based approaches. ### Source excerpt Machine Intelligence [Next page]()