Getting Started with Seer - Sentry's AI Debugging Agent

Getting Started with Seer - Sentry's AI Debugging Agent 🔗 **Source:** [Getting Started with Seer - Sentry's AI Debugging Agent](https://www.youtube.com/watch?v=o6T640JXwUg) > [!note]+ Document Notes > tags: [ai-debugging, sentry, devops] > author: Sentry > source: YouTube > reference: https://www.youtube.com/watch?v=o6T640JXwUg > date created: 2024-12-19 14:30 > > # 🏷️ Title: Getting Started with Seer - Sentry's AI Debugging Agent >  ## 🧰 Action Inbox - [H1] Enable Seer AI debugging agent in Sentry settings — access via left sidebar settings #ai-debugging - [H2] Configure auto-trigger fixes and PR creation — highly actionable issues get automatic fixes #automation - [H3] Set up Sentry MCP server for IDE integration — query application performance from code editor #integration ## 🧭 Table of Contents - [📝 Summary](#-summary) - [📖 Main Content](#-main-content) - [🗒️ Notes](#️-notes) - [📚 References](#-references) - [🗣️ Paraphrase](#️-paraphrase) - [📰 Newsletter](#-newsletter) - [✅ Action Items](#-action-items) - [🧪 Hypotheses](#-hypotheses) - [📊 Methodology](#-methodology) - [❓ Questions](#-questions) - [🔮 Future Research](#-future-research) - [💡 Implications](#-implications) - [🧠 Learning Styles Assessment](#-learning-styles-assessment) - [📊 Accessibility Assessment](#-accessibility-assessment) - [🌍 DEIB Assessment](#-deib-assessment) - [✅ Logical Validity](#-logical-validity) - [🛡️ DARVO Analysis](#-darvo-analysis) - [🎯 Conclusion](#-conclusion) ### 🔗 References - [Sentry Platform](https://sentry.io) - [Seer AI Debugging Agent](https://sentry.io/seer) ## 📝 Summary - **What happened**: Sentry released Seer, an AI debugging agent that provides automated bug detection and fixing - **Why it matters**: Reduces time from bug detection to resolution by automating root cause analysis and fix generation - **Key stakeholders**: Software developers, DevOps teams, engineering managers using Sentry for application monitoring - **Risks and tradeoffs**: Dependency on AI accuracy for automated fixes, potential for false positives in bug detection - **Bottom line takeaway**: Seer integrates AI-powered debugging throughout the entire development workflow from PR review to production monitoring ## 📖 Main Content Seer represents Sentry's entry into AI-powered debugging, combining their existing application monitoring capabilities with artificial intelligence to automate bug detection and resolution. The agent operates across three key phases of development: code review (predicting bugs before production), production monitoring (automatically diagnosing issues), and active development (providing performance insights within IDEs). The system leverages Sentry's comprehensive data collection including distributed tracing, stack traces, logs, profiles, error messages, and codebase access. A reasoning engine processes this context to understand application behavior patterns and predict potential issues. Key capabilities include automated bug prediction during PR reviews, root cause analysis for production issues, automatic fix generation and PR creation, and IDE integration through MCP servers for real-time application insights. ## 🗒️ Notes #### ⏱️ [00:00](https://youtu.be/o6T640JXwUg?t=0) — [H1] Introduction to Seer AI debugging agent - Key idea: Seer is Sentry's AI debugging agent with access to comprehensive application context - Evidence: "It has access to all the context that Sentry collects about the applications that you're building" #### ⏱️ [00:15](https://youtu.be/o6T640JXwUg?t=15) — [H2] Data sources and reasoning capabilities - Key idea: Combines distributed tracing, stack traces, logs, profiles with reasoning engine - Evidence: "It combines all that together and puts a reasoning engine on top of it" #### ⏱️ [00:35](https://youtu.be/o6T640JXwUg?t=35) — [H2] Pre-production bug prediction - Key idea: Analyzes pull requests to predict bugs before production deployment - Evidence: "When you're pulling PRs, we can use that to go and predict bugs before they actually hit production" #### ⏱️ [00:45](https://youtu.be/o6T640JXwUg?t=45) — [H2] Production issue resolution - Key idea: Automatically identifies root causes and proposes fixes for production issues - Evidence: "We can look at those issues, find the root cause, and actually propose a fix for you" #### ⏱️ [01:15](https://youtu.be/o6T640JXwUg?t=75) — [H2] IDE integration capabilities - Key idea: Works within development environments for performance analysis and bug fixing - Evidence: "You can even use it inside of things like cloud code or open code to dig into the performance" #### ⏱️ [01:35](https://youtu.be/o6T640JXwUg?t=95) — [H3] Context importance for AI debugging - Key idea: AI excels at building apps but struggles with debugging without proper context - Evidence: "AI is really good at going out and building applications, but fixing problems inside of them is really hard when you don't have the context" #### ⏱️ [02:00](https://youtu.be/o6T640JXwUg?t=120) — [H1] Initial setup and configuration - Key idea: Seer appears in Sentry settings with configurable options for automated features - Evidence: "You'll see this when you come into the seer option from in the settings" #### ⏱️ [02:15](https://youtu.be/o6T640JXwUg?t=135) — [H2] Auto-trigger and PR creation settings - Key idea: Configurable automation for highly actionable issues with scoring system - Evidence: "We do a scoring automatically behind the scenes on how much SER thinks it can fix an issue" #### ⏱️ [02:45](https://youtu.be/o6T640JXwUg?t=165) — [H2] Code review automation setup - Key idea: Enable automatic code review for new repositories and projects - Evidence: "This is going to turn on the review as I'm pring changes in" #### ⏱️ [03:15](https://youtu.be/o6T640JXwUg?t=195) — [H2] Project-level configuration - Key idea: Per-project settings allow granular control over Seer features - Evidence: "I can see all these settings on a per project basis inside of my environment" #### ⏱️ [03:45](https://youtu.be/o6T640JXwUg?t=225) — [H1] GitHub PR integration demonstration - Key idea: Seer automatically runs on pull requests and identifies potential bugs - Evidence: "Seer automatically ran against this PR and found a bug as part of it" #### ⏱️ [04:30](https://youtu.be/o6T640JXwUg?t=270) — [H1] Production issue management - Key idea: Issues in production show Seer analysis with root cause and solution - Evidence: "I can see Sears already ran. It's found the root cause of the issue automatically and it's got a solution" #### ⏱️ [05:15](https://youtu.be/o6T640JXwUg?t=315) — [H2] Automated PR generation - Key idea: Seer creates pull requests to implement fixes for identified issues - Evidence: "It's even gone and pulled the PR to go and implement the change that'll fix this problem" #### ⏱️ [05:45](https://youtu.be/o6T640JXwUg?t=345) — [H1] MCP server integration - Key idea: Sentry MCP server enables querying application data from within code editors - Evidence: "I've used a Sentry MCP server to reach out and query into Sentry and to understand more about what's happening" #### ⏱️ [06:15](https://youtu.be/o6T640JXwUg?t=375) — [H2] Performance analysis in development - Key idea: Access to slowest endpoints and performance data during active development - Evidence: "I've asked it for the slowest end points and the MCP has used that same reasoning platform" ## 📚 References - [Sentry Documentation](https://docs.sentry.io) - [Seer AI Agent](https://sentry.io/seer) - [GitHub Integration](https://github.com/marketplace/sentry) ## 🗣️ Paraphrase - Seer is an AI-powered debugging assistant that uses Sentry's comprehensive application monitoring data to predict, diagnose, and fix bugs across the entire development lifecycle - The system operates through three main touchpoints: analyzing pull requests for potential issues, automatically diagnosing production problems, and providing performance insights within development environments - Configuration involves enabling automated features like fix suggestions and PR creation, with granular per-project controls for different team needs and workflows ## 📰 Newsletter 🤖 **Sentry Launches Seer: AI-Powered Debugging That Actually Understands Your App** Debugging just got a major upgrade! Sentry's new AI agent, Seer, doesn't just scan code—it understands your entire application context including traces, logs, and performance data to predict bugs before they hit production and automatically fix issues when they do. **What makes this different?** Unlike generic AI coding assistants, Seer has deep insight into how your application actually runs, making its suggestions contextually relevant and actionable. **Get started today:** Enable Seer in your Sentry settings to start getting automated PR reviews and production issue fixes. **Key features:** - 🔍 Pre-production bug prediction in PRs - 🚨 Automatic root cause analysis for production issues - 🔧 AI-generated fixes with pull requests - 💻 IDE integration for real-time performance insights Perfect for teams looking to reduce time from bug detection to resolution while maintaining code quality standards. Resources: [Sentry Platform](https://sentry.io), [Documentation](https://docs.sentry.io), [GitHub Integration](https://github.com/marketplace/sentry) ## ✅ Action Items #### DevRel - Developer advocacy: Create demo showing Seer integration workflow, measure adoption metrics across different project types - Education: Develop workshops on AI-assisted debugging patterns, office hours for Seer configuration - Developer experience: Document best practices for Seer settings, improve onboarding flow based on user feedback - Developer marketing: Campaign highlighting context-aware debugging advantages, success story content - Community management: Monitor feedback on automated PR creation, survey teams on fix accuracy #### OSPO - License Compliance: Review Seer's code analysis capabilities for license detection, train teams on AI-generated code attribution - Security Management: Establish policies for AI-generated security fixes, review automated vulnerability remediation - Software Inventory: Track Seer-generated dependencies in SBOM, document AI tool usage - Developer Support: Provide training on responsible AI debugging practices, tooling for fix validation - Strategic Alignment: KPIs for AI debugging effectiveness, OKRs around bug resolution time reduction #### Community - Maintainers: Template governance for AI-generated contributions, roadmap input on debugging priorities - Contributors: Onboarding materials for AI-assisted development workflows, first-issue programs with Seer guidance - Sponsors: Impact reports showing debugging efficiency improvements, transparent AI usage budgets ## 🧪 Hypotheses - AI debugging agents with comprehensive application context will reduce mean time to resolution (MTTR) by 40-60% compared to manual debugging - Automated bug prediction in pull requests will decrease production incident frequency by 25-35% within 6 months of adoption - Developer productivity will increase measurably when AI debugging tools integrate directly into existing workflows rather than requiring separate platforms - Teams using context-aware AI debugging will show improved code quality metrics and reduced technical debt accumulation over time ### 📊 Methodology - A/B testing with control groups using traditional debugging versus Seer-assisted workflows, measuring MTTR and bug resolution rates - Longitudinal analysis of pull request defect rates before and after Seer implementation across multiple teams and project types - Developer productivity surveys and time-tracking studies comparing debugging workflows with and without AI assistance ### ❓ Questions - How does the accuracy of AI-generated fixes vary across different programming languages, frameworks, and application architectures? - What governance models work best for teams adopting automated PR creation and fix implementation from AI agents? - How do developers adapt their debugging skills and practices when working alongside AI agents with comprehensive application context? ### 🔮 Future Research - Integration patterns between AI debugging agents and existing DevSecOps toolchains for comprehensive application lifecycle management - Effectiveness of context-aware AI debugging across different software domains (web apps, mobile, embedded systems, ML pipelines) - Developer skill evolution and training needs as AI debugging tools become standard practice in software engineering ### 💡 Implications - **Practice**: Shift from reactive debugging to predictive issue prevention, requiring new workflows for AI-generated fix review and validation - **Policy**: Need for governance around AI-generated code contributions, attribution standards, and automated fix approval processes - **Security**: Enhanced vulnerability detection through comprehensive context analysis, but new risks from automated fix implementation - **Education**: Training requirements for effective AI debugging collaboration, understanding AI reasoning and fix validation - **Funding**: Resource allocation toward AI tooling and infrastructure, potential cost savings from reduced debugging time and faster incident resolution ### 🧠 Learning Styles Assessment Reading level estimate: Flesch-Kincaid Grade Level 10-12 #### 👀 Visual learners - Present: Screenshots of Seer interface, GitHub PR integration, settings configuration - Improve: Architecture diagrams showing data flow, visual comparison of before/after debugging workflows #### 🔊 Aural learners - Present: Video demonstration with narration explaining each step - Improve: Podcast-style discussion of AI debugging benefits, audio summary of key features #### 🗣️ Verbal learners - Present: Clear explanations of technical terms, step-by-step instructions, use cases - Improve: Story-driven examples of debugging scenarios, analogies for AI reasoning process #### 🧩 Kinesthetic learners - Present: Hands-on demonstration, step-by-step setup guide - Improve: Interactive checklist for Seer configuration, sandbox environment for testing #### 🧠 Logical learners - Present: Structured workflow explanation, clear cause-and-effect relationships - Improve: Decision tree for configuration options, metrics showing debugging improvements #### 🧑‍🤝‍🧑 Social learners - Present: Team workflow implications, collaboration features - Improve: Discussion prompts for team adoption planning, peer review processes #### 🧘‍♀️ Solitary learners - Present: Self-guided setup instructions, individual configuration options - Improve: Personal reflection questions about debugging practices, self-assessment worksheets #### 🌿 Naturalistic learners - Present: Application ecosystem context, natural workflow integration - Improve: Analogies to natural problem-solving processes, environmental impact considerations ### 📊 Accessibility Assessment #### Screen reader semantics - Present: Basic heading structure in video description - Improve: Comprehensive alt text for interface elements, ARIA landmarks for settings sections #### Keyboard navigation - Present: Standard web interface navigation - Improve: Documented keyboard shortcuts for Seer features, skip links for settings sections #### Contrast and readability - Present: Standard Sentry interface contrast ratios - Improve: High contrast mode documentation, readability testing for AI-generated content #### Text resizing and reflow - Present: Responsive web interface - Improve: Testing at 200% zoom for all Seer components, reflow validation #### Media and captions - Present: Video demonstration available - Improve: Closed captions for video, downloadable transcript, audio descriptions #### Forms and interactions - Present: Standard form controls for settings - Improve: Clear error messaging for configuration issues, accessible confirmation dialogs #### Standards reference - Map Seer interface components to WCAG 2.2 AA success criteria, ensure AI-generated content meets accessibility standards ### 🌍 DEIB Assessment #### Cultural breadth - Present: English-language demonstration - Improve: Multi-region debugging examples, internationalization considerations for AI fixes #### Inclusive language - Present: Technical terminology with explanations - Improve: Plain-language summaries, avoid AI anthropomorphization #### Neurodiversity support - Present: Structured video format, clear sections - Improve: Content chunking indicators, consistent visual hierarchy #### Representation - Present: Technical focus without identity representation - Improve: Diverse debugging scenarios, inclusive team workflow examples #### Psychological safety - Present: Professional technical --- ## Full Transcript [00:00] Since Seir went available this week for [00:02] everybody, I wanted to do a video and [00:03] show how you actually get set up using [00:05] it inside of the applications you're [00:06] building. But first, what actually is [00:08] Seir? Seir Century's AI debugging agent [00:11] and it has access to all the context [00:12] that Sentry collects about the [00:14] applications that you're building. So [00:16] things like distributed tracing, stack [00:18] traces, your logs, profiles, error [00:21] messages and issues that are coming [00:22] through, even access to your codebase, [00:24] and it combines all that together and [00:26] puts a reasoning engine on top of it [00:27] that the agent can tap into to [00:29] understand more about what's happening [00:31] in your application. So when you're [00:32] pulling PRs, we can use that to go and [00:35] predict bugs before they actually hit [00:37] production. And when things do sneak [00:39] into production and break, we can look [00:40] at those issues, find the root cause, [00:42] and actually propose a fix for you, or [00:44] kick it off to a background agent to go [00:46] and implement that change for you. You [00:48] can even use it inside of things like [00:49] cloud code or open code to dig into the [00:51] performance of your application, or just [00:53] ask it to go out and fix bugs, and it'll [00:56] come back with all that context to make [00:57] it easier to fix. Turns out AI is really [01:00] good at going out and building [01:01] applications, but fixing problems inside [01:03] of them is really hard when you don't [01:04] have the context of what's happening. [01:06] And our whole goal with Sears is to have [01:08] it live right alongside your workflow. [01:09] So you can bring it in at any point and [01:11] have it look at what's happening in your [01:13] application to make it better. So in my [01:15] environment, I'm setting it up at the [01:17] first time and I can see the splash [01:18] screen. You'll see this when you come [01:19] into the seer option from in the [01:21] settings on the lefth hand side. I'm [01:23] going to go ahead and drop into testing [01:24] it out now. And now I'm in the seer [01:26] settings so I can actually go in and [01:27] start configuring it to use in my [01:29] applications. So I've got autot trigger [01:31] fixes on by default and I'm going to go [01:32] ahead and turn it on to automatically [01:34] create PRs for me also. So what this is [01:36] going to do is as it finds these highly [01:38] actionable issues, it's going to go and [01:40] suggest fixes for them. We do a scoring [01:42] automatically behind the scenes on how [01:44] much SER thinks it can fix an issue [01:46] that's coming through. So in my case, [01:48] highly fixable ones, those are going to [01:50] get triggered. I'm going to go ahead and [01:52] turn on enable code review by default. [01:53] So this is going to turn on the review [01:55] as I'm pring changes in. So it'll [01:57] actually check those out. So anytime I [01:59] create new repos or projects, this will [02:00] be turned on by default. And then if I [02:02] go into projects here, I can see all [02:04] these settings on a per project basis [02:06] inside of my environment. So if I wanted [02:07] to turn these settings off for specific [02:09] ones, I can. Or if I wanted to enable it [02:11] for other ones that hadn't been turned [02:12] on before, I can do that. Also, if you [02:14] come in and set up a background agent, [02:16] it'll be available here, and you can go [02:17] into the project individually and turn [02:19] those on. So what does this actually [02:20] look like in practice? If I drop into [02:22] GitHub, I'm inside of a PR for one of my [02:24] projects here, and Seer automatically [02:26] ran against this PR and found a bug as [02:28] part of it. So we look at issues as [02:30] they're coming in and we use all the [02:31] context from Sentry to help figure out [02:33] if there's going to be a bug inside of [02:35] the changes you're making as part of [02:36] that code review. Here I was able to see [02:38] what the bug was and I even get a prompt [02:40] that I can feed into like a cloud code [02:42] or a codeex to go and actually implement [02:44] this change for me. I was able to use [02:46] this and send it back in and fix this [02:48] issue before it even hit prod. Now no [02:49] matter how careful we are in PR review, [02:52] things will still make it into [02:53] production that cause issues. And so [02:54] when I jump back into Sentry and I go [02:56] back into my issues view, I'm in a [02:58] project for an iOS application I made. [02:59] And this has a bunch of issues that have [03:01] shown up in it. I can see a couple of [03:03] these have this icon here that show that [03:04] Sears already ran on these. And if I [03:06] jump into these on the right hand side, [03:08] I can see Sears already ran. It's found [03:10] the root cause of the issue [03:11] automatically and it's got a solution to [03:13] go and fix it. It's even gone and pulled [03:14] the PR to go and implement the change [03:16] that'll fix this problem. And that's [03:18] part of what makes Sears so great is it [03:20] lives right alongside your application. [03:21] And as these issues are coming through, [03:23] it's finding them and automatically [03:25] telling you what to do to go and fix [03:26] them. Now, so far, what we've focused on [03:28] is how we're taking code from actually [03:30] in PR all the way into production. But [03:32] what about when you're actually building [03:34] something new inside your editor? I'm [03:35] going to drop down into my terminal [03:37] where I'm in open code right now, and [03:38] I've used a Sentry MCP server to reach [03:41] out and query into Sentry and to [03:43] understand more about what's happening [03:44] inside of my application. I've asked it [03:45] for the slowest end points and the MCP [03:48] has used that same reasoning platform to [03:50] go and pull together all the information [03:52] about what's happening inside of my app. [03:54] So again, slow API calls. I get a nice [03:56] summary of all of those here. I even get [03:57] a dashboard link out to go and dig into [03:59] that inside of Sentry if I want to [04:01] explore. This lets us expose all the [04:03] information that Century has about how [04:05] our application's been running as part [04:06] of the development flow also. So whether [04:08] I'm in development, I'm in review, or [04:11] I'm in production, Sears there helping [04:13] me go and fix things before they start [04:14] to become bigger problems in my [04:16] environment. And that's really the whole [04:17] idea behind Sears. We wanted to be that [04:19] agent that lives right alongside your [04:21] existing development workflow and helps [04:23] you fix problems inside of your [04:24] applications before they become bigger [04:26] ones. Let us know how you want to use [04:27] Seer and how it's showing up inside of [04:29] your workflows. We'll talk to you soon.  ## 📺 Video ![](https://www.youtube.com/watch?v=o6T640JXwUg) --- # 📝 Full Notes (Deep-Read Analysis) # 🏷️ Title: Getting Started with Seer - Sentry's AI Debugging Agent ## 🧰 Action Inbox - [H1] Enable Seer AI debugging agent in Sentry settings — access via left sidebar settings #ai-debugging - [H2] Configure auto-trigger fixes and PR creation — highly actionable issues get automatic fixes #automation - [H3] Set up Sentry MCP server for IDE integration — query application performance from code editor #integration ## 🧭 Table of Contents - [📝 Summary](#-summary) - [📖 Main Content](#-main-content) - [🗒️ Notes](#️-notes) - [📚 References](#-references) - [🗣️ Paraphrase](#️-paraphrase) - [📰 Newsletter](#-newsletter) - [✅ Action Items](#-action-items) - [🧪 Hypotheses](#-hypotheses) - [📊 Methodology](#-methodology) - [❓ Questions](#-questions) - [🔮 Future Research](#-future-research) - [💡 Implications](#-implications) - [🧠 Learning Styles Assessment](#-learning-styles-assessment) - [📊 Accessibility Assessment](#-accessibility-assessment) - [🌍 DEIB Assessment](#-deib-assessment) - [✅ Logical Validity](#-logical-validity) - [🛡️ DARVO Analysis](#-darvo-analysis) - [🎯 Conclusion](#-conclusion) ### 🔗 References - [Sentry Platform](https://sentry.io) - [Seer AI Debugging Agent](https://sentry.io/seer) ## 📝 Summary - **What happened**: Sentry released Seer, an AI debugging agent that provides automated bug detection and fixing - **Why it matters**: Reduces time from bug detection to resolution by automating root cause analysis and fix generation - **Key stakeholders**: Software developers, DevOps teams, engineering managers using Sentry for application monitoring - **Risks and tradeoffs**: Dependency on AI accuracy for automated fixes, potential for false positives in bug detection - **Bottom line takeaway**: Seer integrates AI-powered debugging throughout the entire development workflow from PR review to production monitoring ## 📖 Main Content Seer represents Sentry's entry into AI-powered debugging, combining their existing application monitoring capabilities with artificial intelligence to automate bug detection and resolution. The agent operates across three key phases of development: code review (predicting bugs before production), production monitoring (automatically diagnosing issues), and active development (providing performance insights within IDEs). The system leverages Sentry's comprehensive data collection including distributed tracing, stack traces, logs, profiles, error messages, and codebase access. A reasoning engine processes this context to understand application behavior patterns and predict potential issues. Key capabilities include automated bug prediction during PR reviews, root cause analysis for production issues, automatic fix generation and PR creation, and IDE integration through MCP servers for real-time application insights. ## 🗒️ Notes #### ⏱️ [00:00](https://youtu.be/o6T640JXwUg?t=0) — [H1] Introduction to Seer AI debugging agent - Key idea: Seer is Sentry's AI debugging agent with access to comprehensive application context - Evidence: "It has access to all the context that Sentry collects about the applications that you're building" #### ⏱️ [00:15](https://youtu.be/o6T640JXwUg?t=15) — [H2] Data sources and reasoning capabilities - Key idea: Combines distributed tracing, stack traces, logs, profiles with reasoning engine - Evidence: "It combines all that together and puts a reasoning engine on top of it" #### ⏱️ [00:35](https://youtu.be/o6T640JXwUg?t=35) — [H2] Pre-production bug prediction - Key idea: Analyzes pull requests to predict bugs before production deployment - Evidence: "When you're pulling PRs, we can use that to go and predict bugs before they actually hit production" #### ⏱️ [00:45](https://youtu.be/o6T640JXwUg?t=45) — [H2] Production issue resolution - Key idea: Automatically identifies root causes and proposes fixes for production issues - Evidence: "We can look at those issues, find the root cause, and actually propose a fix for you" #### ⏱️ [01:15](https://youtu.be/o6T640JXwUg?t=75) — [H2] IDE integration capabilities - Key idea: Works within development environments for performance analysis and bug fixing - Evidence: "You can even use it inside of things like cloud code or open code to dig into the performance" #### ⏱️ [01:35](https://youtu.be/o6T640JXwUg?t=95) — [H3] Context importance for AI debugging - Key idea: AI excels at building apps but struggles with debugging without proper context - Evidence: "AI is really good at going out and building applications, but fixing problems inside of them is really hard when you don't have the context" #### ⏱️ [02:00](https://youtu.be/o6T640JXwUg?t=120) — [H1] Initial setup and configuration - Key idea: Seer appears in Sentry settings with configurable options for automated features - Evidence: "You'll see this when you come into the seer option from in the settings" #### ⏱️ [02:15](https://youtu.be/o6T640JXwUg?t=135) — [H2] Auto-trigger and PR creation settings - Key idea: Configurable automation for highly actionable issues with scoring system - Evidence: "We do a scoring automatically behind the scenes on how much SER thinks it can fix an issue" #### ⏱️ [02:45](https://youtu.be/o6T640JXwUg?t=165) — [H2] Code review automation setup - Key idea: Enable automatic code review for new repositories and projects - Evidence: "This is going to turn on the review as I'm pring changes in" #### ⏱️ [03:15](https://youtu.be/o6T640JXwUg?t=195) — [H2] Project-level configuration - Key idea: Per-project settings allow granular control over Seer features - Evidence: "I can see all these settings on a per project basis inside of my environment" #### ⏱️ [03:45](https://youtu.be/o6T640JXwUg?t=225) — [H1] GitHub PR integration demonstration - Key idea: Seer automatically runs on pull requests and identifies potential bugs - Evidence: "Seer automatically ran against this PR and found a bug as part of it" #### ⏱️ [04:30](https://youtu.be/o6T640JXwUg?t=270) — [H1] Production issue management - Key idea: Issues in production show Seer analysis with root cause and solution - Evidence: "I can see Sears already ran. It's found the root cause of the issue automatically and it's got a solution" #### ⏱️ [05:15](https://youtu.be/o6T640JXwUg?t=315) — [H2] Automated PR generation - Key idea: Seer creates pull requests to implement fixes for identified issues - Evidence: "It's even gone and pulled the PR to go and implement the change that'll fix this problem" #### ⏱️ [05:45](https://youtu.be/o6T640JXwUg?t=345) — [H1] MCP server integration - Key idea: Sentry MCP server enables querying application data from within code editors - Evidence: "I've used a Sentry MCP server to reach out and query into Sentry and to understand more about what's happening" #### ⏱️ [06:15](https://youtu.be/o6T640JXwUg?t=375) — [H2] Performance analysis in development - Key idea: Access to slowest endpoints and performance data during active development - Evidence: "I've asked it for the slowest end points and the MCP has used that same reasoning platform" ## 📚 References - [Sentry Documentation](https://docs.sentry.io) - [Seer AI Agent](https://sentry.io/seer) - [GitHub Integration](https://github.com/marketplace/sentry) ## 🗣️ Paraphrase - Seer is an AI-powered debugging assistant that uses Sentry's comprehensive application monitoring data to predict, diagnose, and fix bugs across the entire development lifecycle - The system operates through three main touchpoints: analyzing pull requests for potential issues, automatically diagnosing production problems, and providing performance insights within development environments - Configuration involves enabling automated features like fix suggestions and PR creation, with granular per-project controls for different team needs and workflows ## 📰 Newsletter 🤖 **Sentry Launches Seer: AI-Powered Debugging That Actually Understands Your App** Debugging just got a major upgrade! Sentry's new AI agent, Seer, doesn't just scan code—it understands your entire application context including traces, logs, and performance data to predict bugs before they hit production and automatically fix issues when they do. **What makes this different?** Unlike generic AI coding assistants, Seer has deep insight into how your application actually runs, making its suggestions contextually relevant and actionable. **Get started today:** Enable Seer in your Sentry settings to start getting automated PR reviews and production issue fixes. **Key features:** - 🔍 Pre-production bug prediction in PRs - 🚨 Automatic root cause analysis for production issues - 🔧 AI-generated fixes with pull requests - 💻 IDE integration for real-time performance insights Perfect for teams looking to reduce time from bug detection to resolution while maintaining code quality standards. Resources: [Sentry Platform](https://sentry.io), [Documentation](https://docs.sentry.io), [GitHub Integration](https://github.com/marketplace/sentry) ## ✅ Action Items #### DevRel - Developer advocacy: Create demo showing Seer integration workflow, measure adoption metrics across different project types - Education: Develop workshops on AI-assisted debugging patterns, office hours for Seer configuration - Developer experience: Document best practices for Seer settings, improve onboarding flow based on user feedback - Developer marketing: Campaign highlighting context-aware debugging advantages, success story content - Community management: Monitor feedback on automated PR creation, survey teams on fix accuracy #### OSPO - License Compliance: Review Seer's code analysis capabilities for license detection, train teams on AI-generated code attribution - Security Management: Establish policies for AI-generated security fixes, review automated vulnerability remediation - Software Inventory: Track Seer-generated dependencies in SBOM, document AI tool usage - Developer Support: Provide training on responsible AI debugging practices, tooling for fix validation - Strategic Alignment: KPIs for AI debugging effectiveness, OKRs around bug resolution time reduction #### Community - Maintainers: Template governance for AI-generated contributions, roadmap input on debugging priorities - Contributors: Onboarding materials for AI-assisted development workflows, first-issue programs with Seer guidance - Sponsors: Impact reports showing debugging efficiency improvements, transparent AI usage budgets ## 🧪 Hypotheses - AI debugging agents with comprehensive application context will reduce mean time to resolution (MTTR) by 40-60% compared to manual debugging - Automated bug prediction in pull requests will decrease production incident frequency by 25-35% within 6 months of adoption - Developer productivity will increase measurably when AI debugging tools integrate directly into existing workflows rather than requiring separate platforms - Teams using context-aware AI debugging will show improved code quality metrics and reduced technical debt accumulation over time ### 📊 Methodology - A/B testing with control groups using traditional debugging versus Seer-assisted workflows, measuring MTTR and bug resolution rates - Longitudinal analysis of pull request defect rates before and after Seer implementation across multiple teams and project types - Developer productivity surveys and time-tracking studies comparing debugging workflows with and without AI assistance ### ❓ Questions - How does the accuracy of AI-generated fixes vary across different programming languages, frameworks, and application architectures? - What governance models work best for teams adopting automated PR creation and fix implementation from AI agents? - How do developers adapt their debugging skills and practices when working alongside AI agents with comprehensive application context? ### 🔮 Future Research - Integration patterns between AI debugging agents and existing DevSecOps toolchains for comprehensive application lifecycle management - Effectiveness of context-aware AI debugging across different software domains (web apps, mobile, embedded systems, ML pipelines) - Developer skill evolution and training needs as AI debugging tools become standard practice in software engineering ### 💡 Implications - **Practice**: Shift from reactive debugging to predictive issue prevention, requiring new workflows for AI-generated fix review and validation - **Policy**: Need for governance around AI-generated code contributions, attribution standards, and automated fix approval processes - **Security**: Enhanced vulnerability detection through comprehensive context analysis, but new risks from automated fix implementation - **Education**: Training requirements for effective AI debugging collaboration, understanding AI reasoning and fix validation - **Funding**: Resource allocation toward AI tooling and infrastructure, potential cost savings from reduced debugging time and faster incident resolution ### 🧠 Learning Styles Assessment Reading level estimate: Flesch-Kincaid Grade Level 10-12 #### 👀 Visual learners - Present: Screenshots of Seer interface, GitHub PR integration, settings configuration - Improve: Architecture diagrams showing data flow, visual comparison of before/after debugging workflows #### 🔊 Aural learners - Present: Video demonstration with narration explaining each step - Improve: Podcast-style discussion of AI debugging benefits, audio summary of key features #### 🗣️ Verbal learners - Present: Clear explanations of technical terms, step-by-step instructions, use cases - Improve: Story-driven examples of debugging scenarios, analogies for AI reasoning process #### 🧩 Kinesthetic learners - Present: Hands-on demonstration, step-by-step setup guide - Improve: Interactive checklist for Seer configuration, sandbox environment for testing #### 🧠 Logical learners - Present: Structured workflow explanation, clear cause-and-effect relationships - Improve: Decision tree for configuration options, metrics showing debugging improvements #### 🧑‍🤝‍🧑 Social learners - Present: Team workflow implications, collaboration features - Improve: Discussion prompts for team adoption planning, peer review processes #### 🧘‍♀️ Solitary learners - Present: Self-guided setup instructions, individual configuration options - Improve: Personal reflection questions about debugging practices, self-assessment worksheets #### 🌿 Naturalistic learners - Present: Application ecosystem context, natural workflow integration - Improve: Analogies to natural problem-solving processes, environmental impact considerations ### 📊 Accessibility Assessment #### Screen reader semantics - Present: Basic heading structure in video description - Improve: Comprehensive alt text for interface elements, ARIA landmarks for settings sections #### Keyboard navigation - Present: Standard web interface navigation - Improve: Documented keyboard shortcuts for Seer features, skip links for settings sections #### Contrast and readability - Present: Standard Sentry interface contrast ratios - Improve: High contrast mode documentation, readability testing for AI-generated content #### Text resizing and reflow - Present: Responsive web interface - Improve: Testing at 200% zoom for all Seer components, reflow validation #### Media and captions - Present: Video demonstration available - Improve: Closed captions for video, downloadable transcript, audio descriptions #### Forms and interactions - Present: Standard form controls for settings - Improve: Clear error messaging for configuration issues, accessible confirmation dialogs #### Standards reference - Map Seer interface components to WCAG 2.2 AA success criteria, ensure AI-generated content meets accessibility standards ### 🌍 DEIB Assessment #### Cultural breadth - Present: English-language demonstration - Improve: Multi-region debugging examples, internationalization considerations for AI fixes #### Inclusive language - Present: Technical terminology with explanations - Improve: Plain-language summaries, avoid AI anthropomorphization #### Neurodiversity support - Present: Structured video format, clear sections - Improve: Content chunking indicators, consistent visual hierarchy #### Representation - Present: Technical focus without identity representation - Improve: Diverse debugging scenarios, inclusive team workflow examples #### Psychological safety - Present: Professional technical --- # 📄 Full Transcript *Auto-generated from YouTube captions · timestamps roughly every 30 seconds · [watch on YouTube](https://www.youtube.com/watch?v=o6T640JXwUg)* **[00:00]** Since Seir went available this week for everybody, I wanted to do a video and show how you actually get set up using it inside of the applications you're building. But first, what actually is Seir? Seir Century's AI debugging agent and it has access to all the context that Sentry collects about the applications that you're building. So things like distributed tracing, stack traces, your logs, profiles, error messages and issues that are coming through, even access to your codebase, and it combines all that together and puts a reasoning engine on top of it that the agent can tap into to understand more about what's happening **[00:31]** in your application. So when you're pulling PRs, we can use that to go and predict bugs before they actually hit production. And when things do sneak into production and break, we can look at those issues, find the root cause, and actually propose a fix for you, or kick it off to a background agent to go and implement that change for you. You can even use it inside of things like cloud code or open code to dig into the performance of your application, or just ask it to go out and fix bugs, and it'll come back with all that context to make it easier to fix. Turns out AI is really good at going out and building **[01:01]** applications, but fixing problems inside of them is really hard when you don't have the context of what's happening. And our whole goal with Sears is to have it live right alongside your workflow. So you can bring it in at any point and have it look at what's happening in your application to make it better. So in my environment, I'm setting it up at the first time and I can see the splash screen. You'll see this when you come into the seer option from in the settings on the lefth hand side. I'm going to go ahead and drop into testing it out now. And now I'm in the seer settings so I can actually go in and start configuring it to use in my applications. So I've got autot trigger fixes on by default and I'm going to go **[01:32]** ahead and turn it on to automatically create PRs for me also. So what this is going to do is as it finds these highly actionable issues, it's going to go and suggest fixes for them. We do a scoring automatically behind the scenes on how much SER thinks it can fix an issue that's coming through. So in my case, highly fixable ones, those are going to get triggered. I'm going to go ahead and turn on enable code review by default. So this is going to turn on the review as I'm pring changes in. So it'll actually check those out. So anytime I create new repos or projects, this will be turned on by default. And then if I go into projects here, I can see all **[02:04]** these settings on a per project basis inside of my environment. So if I wanted to turn these settings off for specific ones, I can. Or if I wanted to enable it for other ones that hadn't been turned on before, I can do that. Also, if you come in and set up a background agent, it'll be available here, and you can go into the project individually and turn those on. So what does this actually look like in practice? If I drop into GitHub, I'm inside of a PR for one of my projects here, and Seer automatically ran against this PR and found a bug as part of it. So we look at issues as they're coming in and we use all the context from Sentry to help figure out if there's going to be a bug inside of **[02:35]** the changes you're making as part of that code review. Here I was able to see what the bug was and I even get a prompt that I can feed into like a cloud code or a codeex to go and actually implement this change for me. I was able to use this and send it back in and fix this issue before it even hit prod. Now no matter how careful we are in PR review, things will still make it into production that cause issues. And so when I jump back into Sentry and I go back into my issues view, I'm in a project for an iOS application I made. And this has a bunch of issues that have shown up in it. I can see a couple of these have this icon here that show that Sears already ran on these. And if I **[03:06]** jump into these on the right hand side, I can see Sears already ran. It's found the root cause of the issue automatically and it's got a solution to go and fix it. It's even gone and pulled the PR to go and implement the change that'll fix this problem. And that's part of what makes Sears so great is it lives right alongside your application. And as these issues are coming through, it's finding them and automatically telling you what to do to go and fix them. Now, so far, what we've focused on is how we're taking code from actually in PR all the way into production. But what about when you're actually building something new inside your editor? I'm going to drop down into my terminal **[03:37]** where I'm in open code right now, and I've used a Sentry MCP server to reach out and query into Sentry and to understand more about what's happening inside of my application. I've asked it for the slowest end points and the MCP has used that same reasoning platform to go and pull together all the information about what's happening inside of my app. So again, slow API calls. I get a nice summary of all of those here. I even get a dashboard link out to go and dig into that inside of Sentry if I want to explore. This lets us expose all the information that Century has about how our application's been running as part of the development flow also. So whether **[04:08]** I'm in development, I'm in review, or I'm in production, Sears there helping me go and fix things before they start to become bigger problems in my environment. And that's really the whole idea behind Sears. We wanted to be that agent that lives right alongside your existing development workflow and helps you fix problems inside of your applications before they become bigger ones. Let us know how you want to use Seer and how it's showing up inside of your workflows. We'll talk to you soon.