Conference talk by Mingsheng Hong (Ironclad) on managing AI token usage through outcome-oriented metrics rather than cost leaderboards. Argues against optimizing token consumption directly — comparing it to lines of code as a metric worth tracking but harmful to optimize — and instead measures value alongside cost. Covers the evolution of the team's value metric from lines of code to open pull requests, merged pull requests, and finally complexity-weighted merges; the concept of "trusted throughput" (work passing objective checks, human review, and customer contact); how the bottleneck shifts to code review and CI, encouraging hard-to-review batched pull requests; and practical remedies such as fixing flaky tests, capping agent retry loops, and measuring wait time from ready to merged. Timestamps indicate additional topics including multi-vendor cost measurement, AI as first pass with humans for judgment, guardrails and anomaly alerts, prompt caching and context pruning, and build-vs-buy decisions. # Channel Presentation Observations: "From Tokenmaxxing to Trusted Throughput — Mingsheng Hong, Ironclad" ## Explicit Self-Description - **Speaker role stated**: "I'm Mingshan, VP of engineering focused on AI at Ironclad." - **Company description (creator claim)**: "ironclad is a legal contracting AI companies AI company. We build AI features and native AI products to help lawyers, procurement and other business users move forward new contracts, move them forward faster with controlled risk." - **Stated company priority**: "building trust is the number one priority with our AI product features and products." - **Hiring mention**: "we are hiring... if you know of someone who is interested in building cutting edge legal contracting AI we would love to talk." ## Intended Audience - **Engineering leaders at companies already past initial AI adoption** — stated directly: > "I want to make sure that we position this talk for those of you whose teams have already gone through the hump of getting AI adopted." > "This is a very topical subject that every engineering leader I believe is navigating." - Confirmed by the closing line: "I know this is engineering leadership crowd." - Audience interaction is present ("Can I get a quick show of hands..."), indicating a **live conference/meetup talk** rather than a studio production. ## Claimed Scope of Content - Topic framed at the outset: "how to control and optimize for your AI token spend." - Introduces a **framework the speaker claims as their team's proposal**: "we propose a concept we call trusted throughput." - Positioning claim: "The goal is not to minimizing or not even necessarily to reduce token spend... it's about further improving the ROI of the token spend." ## Observable Presentation Patterns - **Practitioner-experience format**: shares internal practices ("what we at ironclad does"), lessons learned ("we just got over that hump over the last couple quarters"), and evolving metrics ("that's still something we're going to evolve"). - **Cross-references industry anecdotes** (Amazon token-usage dashboard story, Meta, "$500 million on cloud" story) without claiming them as Ironclad experience. - **Cautions/anti-patterns offered explicitly**: leaderboard-style token dashboards, submitting large PRs to bypass CI load, unbounded agentic loops. - **Hedging on maturity**: acknowledges open questions — "I would love to discuss with everyone on kind of how we end up creating defining a set of metrics" and "sometimes it's still a bit ambiguous" on build-vs-buy. - **References a prior speaker**: "for the prior speaker, she did a wonderful job telling you about the importance of trust," confirming this is one segment of a larger event. ## Evidence Notes - All observations above are drawn from direct transcript statements; **no curated metadata was supplied** for this video. - No claims were found about the channel itself (e.g., subscriber counts, channel mission); identity evidence is limited to the speaker's, Ironclad's, and the event audience's descriptions within the talk. # Delivery Observations: "From Tokenmaxxing to Trusted Throughput" — Mingsheng Hong ## Tone & Register - **Conversational, collegial, engineering-leader register** — frequent fillers and hedges ("uh," "you know," "kind of," "right") throughout. > "So, ironclad is a legal contracting AI companies AI company." > "So I assume all of you deploy some form of CI uh CI/CD" - Consistently **inclusive framing** ("we," "us," "you all") rather than prescriptive lecturing: > "that's how we kind of close the learning loop." ## Recurring Phrases & Verbal Habits - **"kind of"** as a recurring softener — appears dozens of times: > "kind of stack rank people, right?" > "just took a pragmatic approach" > "kind of a voluntary dashboard" - **"right?"** used as a check-in tag: > "start reducing or minimizing the cost, right? Cutting cost." > "you know, R&D work to try out things without lending them" - **"So"** as the dominant transition word, opening most new points: > "So how do we think about the policies?" > "So let's talk about then how we can control..." > "So with all of the analysis and ideas here..." - **"I want to share / I wanted to share"** as a recurring framing device: > "one initial learning I want to share is..." > "I wanted to share with you a bit more about what we at ironclad does" ## Teaching Habits & Structure - **Live audience polling** — a repeated pattern (at least twice): > "Can I get a get a quick show of hands that this is a relevant topic?" > "Can I get a quick show of hand for those of you whose teams have gone over the initial adoption phase... I see roughly half of the hands raised." - **Analogies to familiar concepts** as explanation scaffolding: > "We think of the the usage dashboard more as a smoke detector." > "There's an interesting analogy I want to draw with uh a traditional edge productivity metric called lines of code." - **Numbered / bucketed taxonomies** for organization: > "So these are the three buckets that together form what we think is trusted throughput." > "It has three aspects. The first one is..." - **Evolutionary framing** for metrics — presenting each step then critiquing it: > "So let's measure the open PR... But the next question is not every merged PR is equal." > "So that's an improvement. But the next question is..." - **Explicit signposting / roadmap-giving**: > "In the next couple slides, I'm going to introduce two new bottlenecks..." > "Now let's look at the qualitative view." - **Labelling anti-patterns explicitly**: > "so that is an anti-attern I wanted to caution" > "one pitfall I have seen and we wanted to caution everybody" - **Quoting hypothetical stakeholders** in dialogue form to ground points: > "people say, 'Hey, I used to really take pride and joy in handcrafting the code and now a lot of the joy and the pride got taken away'" > "a workar around by engineers... 'I don't want to break my PR into 10 right which might take 10 hours'" ## Audience Awareness & Dialogue Posture - Repeatedly **checks resonance with the room**: > "Does that resonate? I see some heads nodding." - Frames the talk as the **start of a dialogue**, not a finished answer: > "I would love to start that dialogue with you all today to explore the best practices." > "I would love to discuss with everyone on kind of how we end up creating defining a set of metrics..." - **Pre-qualifies the audience** explicitly at the start: > "I want to make sure that we position this talk for those of you whose teams have already gone through the hump of getting AI adopted." ## Honest Uncertainty / Openness - Willingly flags **unresolved questions** and secondhand anecdotes: > "I'm not sure there's explicit encouragement from the leadership..." > "There's no traditional definition of what that means." > "sometimes it's still a bit ambiguous..." - Acknowledges his own organization's recency: > "frankly, we just got over that hump over the last couple quarters." ## Speaking Imperfections (Observable, Not Judgments) - Notable **false starts and self-corrections**: > "There can be a PO with only 10 lines" (likely "PR") > "anti-attern anti-solution" / "anti-attern" > "for the prior speak uh speaker speaker, she did a wonderful job" > "the the usage dashboard" - Occasional **run-on sentences** stitched with "and": > "when the rubber meets the road how customer perceive the changes are there production fire that lead to ro rollbacks do customers complain have tickets..." ## Uncertainty Notes - Audience laughter, pacing, or visual emphasis cannot be verified from the transcript alone; observations here are limited to linguistic evidence. - The "show of hands" moments and "heads nodding" indicate a live audience, but audience size or reaction intensity is not observable. # Observation Record: "From Tokenmaxxing to Trusted Throughput" — Mingsheng Hong, Ironclad ## Speaker - **Mingsheng Hong**, VP of Engineering focused on AI at Ironclad (self-identified in transcript). ## Framing & Audience Positioning - Topic: controlling and optimizing AI token spend in engineering teams. - Explicitly scoped to teams **past the adoption hump**: those still provisioning access/encouraging adoption "may not be ready" for cost-control ideas (he notes Ironclad itself "just got over that hump over the last couple quarters"). - Stated goal is **not austerity**: "The goal is not to minimizing or not even necessarily to reduce token spend... It's about further improving the ROI of the token spend." ## Opening Anecdotes (as reported by speaker, with qualifications) - **Amazon story**: a voluntary token-usage dashboard led some engineers to compete to top a "leaderboard"; speaker notes "I'm not sure there's explicit encouragement from the leadership." - A similar story from **Meta**, and a "more sensational" claim of companies spending "$500 million on cloud... within a month." - Speaker's qualification: "while these may not be happening in your companies today, the threats, the risks are real." ## Dashboards as "Smoke Detectors," Not Leaderboards - Track per-team, per-individual usage and cost, but **position as a smoke detector**, not a leaderboard. - Low-usage pockets may signal adoption gaps worth investigating. - Explicitly avoid creating "even indirect incentive to maximize the token usage itself." - Analogy to **lines of code (LOC)**: important metric, but unwise to optimize directly — "one can argue that removing code is even better." ## Adoption Lessons (for teams still in that phase) - After top-down leadership pushes, sit down with resistant/struggling individuals. - Quotes a legitimate concern heard from engineers: > "I used to really take pride and joy in handcrafting the code and now a lot of the joy and the pride got taken away and replaced with me reviewing AI slop code." - Implication stated: leadership should identify high-impact technical work where engineers can still grow in the AI era. ## Ironclad Context - Ironclad is a **legal contracting AI company**; building trust is "the number one priority." - Trust-building example: lawyers test conversational search against familiar contracts before expanding to unfamiliar searches, redlining, anomaly detection. - Claim: gaining engineering value from AI similarly requires a sequence of trust-building steps with internal engineers, leadership, and customers. ## Measuring Cost - Vendor dashboards (e.g., Claude Code, Codex) suffice for single-tool shops; Ironclad uses multiple tools and built AI-assisted dashboards/pipelines to cross-correlate and aggregate per-team/per-individual cost. - **Pitfall cautioned against**: jumping from measuring cost to cutting cost — "We think that is premature." ## "Trusted Throughput" — Quantitative Metrics (evolution described) 1. **LOC** — rejected as a value measure. 2. **Open PR count** — shows inflection but insufficient. 3. **Merged PR count** — improvement ("we're all measured by the code we ship"). 4. **Complexity-weighted merged PRs** — pragmatic approach: feed PRs to an LLM with a well-crafted prompt; score complexity by t-shirt size; more complex AI-generated PR treated as more valuable. - Speaker flags this is **still evolving** and open to discussion — an explicit unresolved tension. ## Trusted Throughput — Qualitative Definition Three buckets of validation: - **Objective metrics**: test coverage, predefined security checks, canarying practices. - **Subjective human judgment**: code review, design review — quality, clarity, maintainability, architecture fit. - **External/customer validation**: production fires, rollbacks, customer complaints, tickets about usability/bugs. ## New Bottlenecks: Code Review & CI - Claim: AI makes PR creation abundant, shifting lifecycle pressure onto **code review and CI/merging**. - **Anti-pattern flagged**: if CI is overloaded, engineers stop splitting PRs and submit large ones — risky because it raises human review overhead and thins review attention. - **Code review principle**: AI tooling as "first level of defense" (style issues, missing test coverage), not replacing humans; humans focus on judgment (architecture, security design) and retain "final accountability." - **CI pressure**: smaller/more numerous PRs strain CI; workarounds (manual reruns of flaky tests, AI agents babysitting PRs) "waste AI token" and hurt morale. Response: dev-ex/platform investment in removing flaky tests and improving CI; metrics include clock time from PR-ready to submitted and number of retries per PR. ## Pragmatic Framework (three aspects) 1. **Guardrails**: budgets, quotas, usage tracking, anomaly alerts. 2. **Human review** as complement, feeding learnings into an institutional knowledge base. 3. **Learning loop**: leadership works with individuals to define guardrails, review metrics, refine. ## Best Practices for Token Efficiency - **Agentic loops** in harnesses (e.g., generate PR → run tests → auto-fix → retry), with a **limit on loop steps** to avoid uncontrolled token waste. - **Prompt caching**: structure prompts with fixed system prompt at top, varying content at bottom. - **Context pruning**: summarize long-session contexts; tools like Claude Code increasingly auto-compact — improves both efficiency and output quality. ## Build vs. Buy - **Buy** non-differentiating infrastructure (IDE, CI). - **Build** context-specific assets: internal playbook of well-crafted prompts for PR generation (bug fixes, UI features, refactoring), reused and enhanced across teams. - **Ambiguous case acknowledged**: their "builder agent" (cloud-based code generation wrapping Claude Code) — vendors exist, exploration ongoing; speaker invites exchange of views. ## Closing Claims - Key lessons: learn from others' stories/mistakes; plan ahead for review/CI bottlenecks before scaling code generation; instrument the right metrics for CI health and developer experience. - "We believe that this is the golden era of AI where maximizing token ROI is the key for every team success." - Ends with a hiring pitch for Ironclad's legal contracting AI work. # Speaker Attribution & Provenance Record **Video:** From Tokenmaxxing to Trusted Throughput — Mingsheng Hong, Ironclad **Video ID:** `01a15ca7-99ff-7c76-880d-011032b82dc3` ## Speakers | Speaker | Role | Notes | |---|---|---| | Mingsheng Hong (self-introduced as "Mingshan") | **Creator / Presenter** — VP of Engineering, AI, at Ironclad | Sole identified speaker; delivers the entire talk | - **Referenced prior speaker (unnamed, female):** "the prior speaker... she did a wonderful job telling you about the importance of trust" — identity not given in transcript. - **Audience:** present and interactive (show-of-hands responses, head nodding); not individually identified. ## Promotional Segments - **Closing hiring pitch (flagged):** > "I just want to end with saying we are hiring... if you know of someone who is interested in building cutting edge legal contracting AI we would love to talk." - Ironclad product/company description ("Ironclad is a legal contracting AI company") appears mid-talk as context, not explicitly flagged as promotion by the speaker. ## Referenced Sources & Third-Party Claims - **Amazon anecdote:** media-reported story of an employee-built voluntary dashboard leading engineers to compete on token usage ("tokenmaxxing"); speaker notes uncertainty: *"I'm not sure there's explicit encouragement from the leadership."* - **Meta anecdote:** "a similar story from Meta" — no detail given. - **Unidentified company claim:** "some companies spending $500 million on cloud... within a month" — attributed to media reports. - **Prior speaker's talk** on trust (referenced, not summarized). - **Commercial model vendors:** referenced regarding prompt caching advisories; no vendors named. - **Tools named:** Claude Code ("claw code"), Codex ("codeex") — mentioned as vendor tools/dashboards; no endorsement claims beyond usage description. ## Internal Provenance (Speaker's Own Claims About Ironclad) - Internal dashboards/pipelines built with AI to cross-correlate vendor usage data per team/individual. - Internal PR complexity scoring via AI prompt (t-shirt sizing) — speaker notes "no traditional definition" exists; they "looked at the literature a bit" (unspecified). - Internal AI playbook of reusable prompts; exploratory "builder agent" project. ## Availability & Uncertainty - **Comments:** no transcript or metadata indication of comment availability. - **Metadata:** none supplied beyond title and transcript. - Anecdotal media stories explicitly hedged by speaker ("may not be happening in your companies today"; leadership-encouragement uncertainty at Amazon). All right, let's get started. Apologies for the delay, but I'm really excited to be here. I'm Mingshan, VP of engineering focused on AI at Ironclad. And today I'll be telling you about something that's probably on top of many of your mind. uh how to control and optimize for your AI token spend. Can I get a get a quick show of hands that this is a relevant topic? Okay, awesome. I appreciate that. So, we have all heard a few sensational stories from the media. There's an interesting Amazon story where an employee just created kind of a voluntary dashboard and everyone start tracking their own AI token usage. I'm not sure there's explicit encouragement from the leadership, but the effect is you know engineers some of the engineers started competing with each other in maximizing their token usage and get to the top of the so-called leaderboard. There's a similar story from Meta and then another even more sensational story about some companies spending $500 million on cloud oops within a month. So while these may not be happening in your companies today, the threats, the risks are real. How do we think about the policies? How do we measure the cost? And how do we control and optimize for it? So one initial learning I want to share is it is really important to have dashboard that track every team every individual's token usage and cost but that should not be positioned as a leaderboard. We think of the the usage dashboard more as a smoke detector. If there are local pockets of teams or individuals that don't use much AI token that might be a signal worth investigating. But beyond that certainly we don't want to create even indirect incentive to maximize the token usage itself. So how do we think about it then? First I want to make sure that we position this talk for those of you whose teams have already gone through the hump of getting AI adopted. If you're still in the initial process of provisioning easy access to your engineers or encouraging the teams and individuals to adopt, then you may not be ready to implement some of the ideas for controlling and optimizing for cost. But that's okay. This could still be a good discussion. And frankly, we just got over that hump over the last couple quarters. So this is a very topical subject that every engineering leader I believe is navigating. So I would love to start that dialogue with you all today to explore the best practices. Can I get a quick show of hand for those of you whose teams have gone over the initial adoption phase now you are starting to seriously worry about the cost. Okay, I see roughly half of the hands raised. Thank you. So let's talk about then how we can control and how we can optimize what we call the trusted throughput as a kind of a proxy metric as a way to measure your ROI. But before that, just for those of you who are in the process of still increasing adoption, one lesson we learned is to after the kind of the top down leadership push is to sit down with the individual teams and uh the individuals who may be resistant or struggling with adoption, understand where they came from. For example, there are some legitimate concerns that I heard, you know, people say, "Hey, I used to really take pride and joy in handcrafting the code and now a lot of the joy and the pride got taken away and replaced with me reviewing AI slop code, right? So that doesn't sound like a very satisfying professional activity and that's where we need to kind of dig down and understand what are still the kind of the high impact and uh engineering tasks technical work that we can help our engineers continue to grow themselves in the era of the AI. So I wanted to share with you a bit more about what we at ironclad does and there's an interesting connection actually within how we think about optimizing for engineering AI token usage. So, ironclad is a legal contracting AI companies AI company. We build AI features and native AI products to help lawyers, procurement and other business users move forward new contracts, move them forward faster with controlled risk. What that means is building trust is the number one priority with our AI product features and products. And for the prior speak uh speaker speaker, she did a wonderful job telling you about the importance of trust and how to build it in their domain. In our ironclad product domain, it often means lawyers especially, but other persona as well taking the time to kind of test the water and see if they can trust the AI output. For example, they may feed our conversational search a set of contracts they are firmly familiar with and they run a search and see if the output is towards the expectation. If so, they may expand on searching for things they don't know about or apply other workflows using AI to solve other things like redlinining the contract um and you know finding anomalies and so on. And so similarly using AI and making sure AI is delivering high engineering value also involves a you know a sequence of steps in gaining trust from the internal engineers the leadership as well with as uh with our customers. So this is the focus of our talk today and this probably will not come as a surprise here. The goal is not to minimizing or not even necessarily to reduce token spend. So here we kind of use the word it's not about austerity. It's about further improving the ROI of the token spend. So how do we do that? Here we propose um a concept we call trusted throughput. So the trusted throughput comes from having the code reviewed and validated internally and ultimately validated in customer uh in customer deployments. So how do we go and how do we think about controlling the cost and uh measuring and in turn optimizing the ROI? The first step is I'm pretty confident that all of you your teams who have been adopting AI have been measuring the cost. If you're using a single tool like claw code or codeex then you tend to get very rich analytics from the vendor's dashboard already. If you're like us who use a combination of these different coding tools then we basically use AI to build simple dashboards and pipelines to extract such vendor data. So we can kind of crossorrelate them. Then we can break it down, aggregate and then break down by per team, per individual, what is their cost usage across all of these uh tools. So that's the first step for measuring cost. Now one pitfall I have seen and we wanted to caution everybody is to then jump from measuring cost to start reducing or minimizing the cost, right? Cutting cost. We think that is premature. Instead, the other important side of the equation for ROI is to measure value. How much value are we getting from burning the tokens? Once we can measure the cost and value side, we understand ROI and then to improve ROI, we want to find and then fix the bottlenecks. In the next couple slides, I'm going to introduce two new bottlenecks we identify in this whole new software development life cycle where code generation now becomes abundant thanks to AI. But the pressure is now getting pushed down to code review and continuous integration CICD the merging the code. So we'll talk about that and finally we'll put together these ideas into a pra pragmatic framework of how we think about optimizing the ROI and thus the leverage in using AI. Okay. So this is kind of just a slide in building or using the vendor dashboard to measure the cost. And again we want to caution that here the main goal for regularly reviewing the dashboard is to see a if there's still adoption gap within individual pockets of teams or the individual engineers and b if there are any sudden surprises in kind of the usage burst and if so understand what's been happening if they're legitimate and then also compare teams contextually. So this is important. We don't control just the AI usage per se because for example a platform infrastructure team the way they use AI and the way they get value may be different from the UI team. So we need to take the context into consideration. All of such review analysis is to help us extract learnings. So there's a self-learning loop that we can then feed back into institutional best practices. What we don't want to use the dashboards are to kind of stack rank people, right? making it a a leaderboard and somehow reward maximization. There's an interesting analogy I want to draw with uh a traditional edge productivity metric called lines of code. So I believe all of you will be tracking that metric but it wouldn't be wise to use that metric as the key goal to measure engine velocity because if we want productive and high quality engine work one can argue that removing code is even better. So, LOC line of code is an important metric but not something we want to directly optimize for. Same thing for the token usage and spend. So, that that gets us to the notion of trusted throughput. How do we think about that? How do we define that? First, I want to kind of share the quantified uh side of the things. What are the metrics that kind of we have been involving in defining and tracking. So we talked about line of code is clearly not a good way to measure if AI is you know generating a lot of value. So the next evolution can be let's count the number of open PRs pull requests. The intuition being engineers are using AI to generate a lot more code. So let's measure the open PR. So clearly we see a big kind of inflection in the open PR count. But eventually as we as I assume everyone would agree over the time even though people may do oneoff you know R&D work to try out things without lending them but eventually we're all measured by the code we ship. So therefore we evolved from tracking the open PR count to tracking the merge PR count. So that's an improvement. But the next question is not every merged PR is equal. There can be a PO with only 10 lines of code that takes forever that finds and fix a concurrency bug or there can be a thousand line kind of boilerplate code that just takes a lot of time to then kind of generate and review but otherwise it's not necessary adding as much business value. So as such we then started kind of tagging each merged PR with some sort of complexity score. There's no traditional definition of what that means. We looked at the literature a bit. So we just took a pragmatic approach of giving AI a well-crafted prompt and then we feed the PR into basically one or two M and say score the complexity based on t-shirt size. So I the idea being if you use AI to generate a more complex PR we consider that as being more valuable basically that's how we kind of add a weightage to each merged PR but that's not the end of the journey that's still something we're going to evolve keep evolving and I would love to discuss with everyone on kind of how we end up creating defining a set of metrics that kind of approximate the value AI is generating. Now let's look at the qualitative view. What we think about the way we would define trusted throughput is a high quality output that's interested by both internal engineering and leadership and external customers. We think they come from three buckets. The first bucket is all of the objective metrics that we run with checking the test coverage whether uh all of the predefined security checks are passing. Do we go through the regular canarying practice as we roll out features safely and so on. In addition, we complement the subjective objective metrics with our subjective human judgment. So that's where the code review, the design review come in to look at the code quality, clarity, maintenance, architecture fit and so on. And then finally we want to make sure through all of these internal objective and subjective check when the rubber meets the road how customer perceive the changes are there production fire that lead to ro rollbacks do customers complain have tickets that talk about usability uh friction uh bugs and so on. So these are the three buckets that together form what we think is trusted throughput from engineering. Okay. So now let's talk about from a software deploy deployment life cycle perspective where we observe the new bottlenecks are as I mentioned earlier AI code generation is making PR creation abundant. So now the the bottleneck from kind of the whole life cycle perspective gets shifted onto re review and they're subsequently merging the PR. Does that resonate? I see some heads nodding. So this is where we spend time on figuring out how we can further improve the review process as well as the continuous integration the CI process. So we will dive into these two topics in the next couple slides here. I just want to say a potential anti-attern anti-solution is that hey if the CI infrastructure gets overloaded then a workar around by engineers to stop splitting PR just start submitting large PR for review and submission because if it takes an hour to run all of your regression test and submit it I don't want to break my PR into 10 right which might take 10 hours however this in our view can be pretty risky because it makes the human review overhead higher it also reduce the quality of the review because the human attention can be spread thin so that is an anti-attern I wanted to caution so for code review the key principle we use is to make sure we onboard AI tooling as the first level of defense they don't replace human reviewers but we want to offload human reviewers as much as possible let the AI review take care of simpler things like coding style issues or if there's a missing test coverage. So, make sure the author gets through all of them before then the review gets routed to a human reviewer. And this way our human engineers can focus on applying their deep judgment on aspects that are somewhat subjective like if the code is good, if the architecture is sound, if the code uh uh passes kind of the security uh the security design and so on. so that in the end our engineering team can take the final accountability. Now let's look at CI. So I assume all of you deploy some form of CI uh CI/CD and what we're seeing is thanks to AI now making it much easier to generate code as splitting code into smaller but more PRs it puts a lot more pressure on the CI and this is something that uh if we don't address uh at a company level individual engineers can be struggling because that means they have to waste their human time babysitting the PR to get merged. If they run into flaky test then they have to manually they hit rerun it's very frustrating or they can recruit an AI agent to babysit and kind of do a loop but that in turn waste AI token as well. So these are not these are just workarounds not perfect solution and also tend to make engineers feel a little bit lower morale a little bit more frustrated. So what we what we are doing is kind of we put more uh developer experience uh platform kind of engineering to invest into reducing removing the flaky test improving the CI infrastructure and the key thing here is to also define and measure the right metrics for example uh the work clock time between when a peer is ready to submit till when it's submitted right if a typical CR uh run takes an hour. Does the typical PR submission take two or three hours? In which case, that's a red flag and also the number of times a PR needs to get retrieded for passing through the test. So, these are the key metrics that we are using to measure our developer experiences and the relevant team who is focused on improving uh these uh the developer experience. So with all of the analysis and ideas here we uh want to share kind of the a pragmatic framework of how we can then measure and optimize token usage. It has three aspects. The first one is set the right set of guards across setting the budget and quota tracking usage defining anomalies so that no users uh leaders can get notified if something feels wrong. This is complementaryary to still regular human review which can catch other interesting patterns or learnings and feedback into the institutional knowledge base. Let me just couple that with the third item here which is the learning loop we talk about as our leadership work with individuals to define these guard rails review the metrics and then refine that's how we kind of close the learning loop. In addition to that, we want to work with our teams, individual engineers to continue to search for and if needed innovate on the best practices of how to use AI, how to use AI to build products and also use it internally. For example, some engineers may be writing an agentic loop as part of the harness when they use cloud code. after they generated initial PR they go and loop around and say try and pass the set of tests and then if some tests don't pass just auto fix the test or the code and retry. One thing to watch out for is to put a limit on the number of loop steps to make sure if things go out of control we don't waste too many tokens on that. Another example is prompt caching. This is becoming increasingly more prevalent by the commercial uh model vendors where what they advise is if you send a prompt with the same prefix they could optimize how they process the prefix of the prompt. What that means then as a user to those is that we want to encourage our users to structure their prompt that way. For example, if your prompt consists of a system prompt followed by a user prompt, you want to put the system prompt that's fixed at the top and the varying content at the bottom. Context pruning is also important. We want to kind of drill it into each individual users kind of new kind of muscle memory. So they are aware that as they build out the context through a longer chat session, they would be mindful of summarizing the context and make sure that the token usage is efficient that way. There are increasingly more tools like claw code that will automatically manage and compact the context for you. And so this increases the token usage efficiency but also increase the quality of AI output. There are other ideas we're exploring as well. So I know we're at time so this is towards the end of the talk. There is sometimes we also face build versus by decision. The principle is simple for things that are non- differentiating like IDE CI infrastructure we want to buy. But then for things that are specific to our context like how we would generate high quality PR for small bug fixes versus building a new UI feature for refactoring and so on. We have our internal playbook which is a set of well-crafted AI prompts. So we save that and share across our team. So that gets reused and enhanced. So that's something we must build internally. When it comes to case to case though, sometimes it's still a bit ambiguous like we're trying to build what we call builder agent. That's like a cloud-based code generation that wrap the cloud codec and so on. While we know there are also other vendors out there that we're still exploring. So we love to exchange thoughts on that. So then to summarize here are a couple key lessons as we went through the last couple quarters of journey. I wanted to share so that hopefully you could kind of accelerate your process there. If I were to summarize these three things I would it's about learning planning ahead and learn from other people's stories mistakes. So what that means is think about build respences by early on as you are encouraging more code gen think about how that impact your code review and CI and how you can address these new bottlenecks. And finally, continue to define and instrument your system to get the right metrics to measure the health of your CI system and the whole developer experience in general. So that's it for the talk. We believe that this is the golden era of AI where maximizing token ROI is the key for every team success. And with that, I just want to end with saying we are hiring. I know this is engineering leadership crowd but if you know of someone who is interested in building cutting edge legal contracting AI we would love to talk. Thank you.