Talk recorded at the AI Engineer World's Fair 2026 in San Francisco. Abdul Dakkak, chief scientist at Modular, discusses how the proliferation of AI hardware has strained the software stack, with inference frameworks like vLLM, SGLang, TensorRT-LLM and llama.cpp each working around different hardware. He presents Modular's rebuilt stack: Mojo, a Pythonic systems language targeting CPUs, GPUs and accelerators; open-sourced kernels including matmul and attention; and MAX, an inference framework supporting many model architectures across GPUs, nodes and mixed hardware, with OpenAI-compatible endpoints. Topics include benchmarks (Kimi K2.5, FLUX.2, AMD MI355, Gemma 4), sub-second image generation, and Modular Cloud. Note: performance and benchmark claims are the speaker's own. # Channel Presentation Observations — "Taming the AI Hardware Cambrian Explosion — Abdul Dakkak, Modular" ## Video Type & Setting - This is a **conference/company technical talk**, not a casual channel video. The speaker frames it as a live session: *"today I'm going to be talking about what modular is and what value it brings to you"* and later references *"the booth behind this curtain"* and being *"on the booth for the next few hours."* ## Intended Audience - **AI/ML engineers and infrastructure developers** evaluating inference stacks. Evidence: deep dives into serving frameworks (*"folks actually deploy not only using VLM but also SG lang, TRTLM, Llama CPP"*), benchmark specifics (TTFT, tokens/min per GPU on B200, MI355). - **Developers familiar with Python and PyTorch**: *"follets that actually know Python today pretty much know 90% of Mojo"*; *"If you're familiar with PyTorch programming... will not be a massive departure of your mental model."* - **Prospective hires and open-source contributors**: *"We're hiring... from low-level kernel engineers to high level folks that work on Kubernetes."* ## Claimed Scope - **Full-stack AI platform claim**: Modular positions itself as *"a unifying layer that unifies all the hardware"* — spanning the **Mojo language**, the **Max inference engine**, and the **Modular Cloud**. - **Performance leadership claims**: *"our matrix multiplication is faster than the cuBLAS, the hipBLAS"*; *"5.5x faster on TTFT compared to vLLM"*; *"3x faster than the next competitor"* per Artificial Analysis. - **Openness claims**: *"over 700,000 lines of kernel code that we've open sourced"*; Mojo 1.0 open-sourcing *"in the next few months."* ## Self-Description Evidence - Creator self-descriptions (claims, not independently demonstrated): *"hardware abstracted"*, *"state-of-the-art kernels on Nvidia, AMD and other hardware"*, *"glass box"* cloud, *"independently verified"* benchmarks via Artificial Analysis. ## Demonstrated (vs. Claimed) Presentation Patterns - Heavy use of **comparison framing against incumbents** (CUDA, Triton, OpenCL, PyTorch/vLLM ecosystem), with the channel positioning Modular as filling a *"sweet spot"* of usability, performance, and portability. - Invitation to **verify rather than trust**: *"You can pip install modular right now... I welcome everybody to validate these results."* ## Caveat All quantitative claims are creator- or partner-asserted within the talk; the material demonstrates the channel's pitch and audience posture, not independent verification of the claims. # Delivery Observation — "Taming the AI Hardware Cambrian Explosion" (Abdul Dakkak, Modular) ## Overall Delivery Style - **Conversational, incremental walk-through**: The talk moves layer by layer up the stack, signaled by explicit transitions: > "So let me kind of dig into each one at a time, each part of the stack." > "So the first is Mojo… Sitting on top of Mojo is Max… Finally, layered on top of all of this is the modular cloud." - **Heavy use of "so" as a connective**: It appears in nearly every segment boundary ("So as you probably know…", "So at the bottommost level…", "So how does Perf look like?"). - **Rhetorical questions to pivot sections**: > "But the problem is that you have to ask is how do you program all of these uh systems." > "Now, why do you care about all of this, right? It's a lot of tech." ## Vocabulary & Phrasing Patterns - **Vivid metaphors**: "Cambrian explosion" for hardware proliferation (used in the framing line, "You have Nvidia providing you know a few generations of GPUs, the Intels, the ARMs, the Apples"), hardware choices that "bubble up from the you know from down under uh into upwards into the stack," and the cloud described as "a glass box." - **Superlatives as a recurring device**: "state-of-the-art kernels," "It's blazing fast," "the fastest attention uh kernel out there," "the best way to program CPUs and GPUs." - **Audience-familiarity framing**: Repeated claims that the tools match existing mental models: > "folks that actually know Python today pretty much know 90% of Mojo" > "would or the left will not be like massive departure of your mental model" > "if you're familiar with like any uh cloud provider I mean it's it should be pretty familiar" ## Teaching Habits - **Comparative quadrant/benchmark framing**: Positions Mojo in a "three different quadrants, usability, performance, and portability" and contrasts named competitors (CUDA, OpenCL, Triton) with trade-off language ("they sacrifice on both usability and programmability"). - **Quantified performance claims with named baselines**: "5.5 faster on TTFT uh compared to VLM," "1.5x throughput improvements," "about 7x faster uh compared to ters that compile with diffusers," "4x faster than the equivalent uh on um on on the MI355 systems." - **Caveating as credibility work**: > "you know some of this is not a fair comparison because you know some folks might say oh 7x is much larger than 4x but this is you cannot do that comparison very easily." > "And this is not playing tricks of distillation and you know quantization or anything like that." - **Third-party verification appeals**: "These are these benchmarks are independently verified as well," "artificial analysis actually we've been partnering with them uh and they've independently been verifying our results." ## Audience Engagement Patterns - **Repeated invitations to see the booth/demo**: > "I encourage folks to actually visit the booth behind this curtain, uh there's a nice demo of Flux 2." > "Again you can go to the booth. There's a nice demo…" - **Hands-on call to action near the close**: "You can pip install modular right now if you want to and try all of this stuff. And I welcome everybody to validate these results." - **Explicit "this isn't vaporware" framing**: "So, this is not uh demoare or slideware." ## Disfluencies & Verbal Tics - Frequent filler "uh" throughout and repeated-word stumbles under apparent load: "uh the uh the stack," "as as as you as you as you um as you build," "the max uh the max PF benefit," "never never stopping to improving these things." - Self-corrections mid-sentence: "Uh we're, you know, let's say 6x faster than that." ## Stable vs. Isolated Patterns - **Stable-looking** (recurs across most sections): "so"-led transitions, competitive benchmarking with numbers, booth/demo invitations, familiarity framing for the audience. - **Isolated or thin-evidence moments**: The "Cambrian explosion" metaphor appears once; the "glass box" metaphor once; the fair-comparison caveat appears once (though a similar hedging instinct shows up in the distillation caveat). Confidence in these as habits is low given a single transcript. ## Uncertainty Notes - Transcript lacks slide/diagram content, so claims about visual teaching aids (e.g., "the figure on the right would or the left") cannot be verified. - One word is uncertain in the transcript ("VLM"/"ters" likely represent vLLM and TensorRT), so exact tool-name phrasing is reported as transcribed. # Observation Record — "Taming the AI Hardware Cambrian Explosion" (Abdul Dakkak, Modular) ## Speaker Identification The sole speaker, per title metadata and self-introduction, is **Abdul Dakkak of Modular**. This is a vendor product talk; all performance and capability claims below are his own and were not independently confirmed within the transcript itself (except where he names third-party verification). --- ## Core Claims and Stated Reasoning ### 1. Compute crunch and hardware proliferation - **Claim:** There is a "huge crunch" for compute with "no alleviation in sight," with capacity "booked for the next few years." - **Claim:** This has produced a "**Cambrian explosion**" of hardware — Nvidia GPU generations, Intel, ARM, Apple — each covering niches. - **Stated problem:** "How do you program all of these systems?" Hardware choices "bubble up" into the software stack, making the current software landscape "a mess" (cites vLLM, SGLang, TRT-LLM, Llama.cpp as fragmented deployment options). ### 2. Modular's thesis - **Position:** Rearchitect the stack "from the ground up" around **portability, performance, and programmability**, with a unifying layer on which everything runs. ### 3. Mojo (programming language) - Pythonic heterogeneous systems language, "not unlike C++, Rust or CUDA C++"; knowing Python ≈ "90% of Mojo." - Targets CPUs, GPUs, and accelerators; claimed to "compete with CUDA in terms of performance." - **Positioning framework** (usability / performance / portability quadrants): - CUDA gives massive performance but sacrifices usability and programmability. - OpenCL gave portability at cost of performance and usability. - Triton offers portability/usability but only "70% of the Perf." - Mojo aims for "the best of both worlds." - **Specific performance claims:** - Matrix multiplication faster than cuBLAS and hipBLAS — "not by 1%, in some cases by 2%." - "Fastest attention kernel out there," faster than FlashAttention. - **Announcements:** Mojo 1.0 API stabilization announced ~a month prior; 1.0 release and open-sourcing "in the next few months." ### 4. Max (inference framework) - **Critique of PyTorch:** ~10 years old, "not really designed for the gen AI inference of today"; vLLM/SGLang filled gaps (KV cache, paged attention) but complicated the ecosystem; "a lot of folks don't really use PyTorch as is." - **Claims for Max:** built from the ground up with graph compilers, eager-mode execution, novel Mojo-based fusion; a single cohesive stack for serving, developing, and optimizing. - Supports "thousands of architectures" (naming Gemma 4, MiniMax 3, GLM), OpenAI-compatible endpoints, Python model development, and **native multi-GPU / multi-node programming**, including heterogeneous placement (e.g., prefill on hardware X, decode on hardware Y). ### 5. Modular Cloud - Positioned as a "**glass box**" — all variables in Max/Mojo are tunable within endpoints. - Offers endpoint deployment, bring-your-own-cloud, a cloud lab for exploring the Pareto frontier, and custom model serving (fine-tunes, custom speculative decoders). --- ## Benchmark Claims (speaker-reported) | Claim | Figure | Comparison basis | |---|---|---| | Kimi K2.5 TTFT | 5.5× faster | vs. vLLM | | Kimi K2.5 throughput (B200) | 1.5× tokens/min/GPU | vs. vLLM | | Flux 2 | ~7× faster | vs. "Torch" compiled with Diffusers | | AMD MI355 | 4× faster / TCO benefit | vs. equivalent (unspecified) | | Flux 2 Dev | ~3× faster than next competitor | per Artificial Analysis (named as independent verifier/partner) | | Gemma 4 | ~6× faster | vs. SambaNova (named as a hardware accelerator) | **Qualifications the speaker himself raised:** - The 7× vs. 4× comparisons "cannot [be compared] very easily" because AMD hardware is "usually 30% cheaper" than Nvidia equivalents — so raw multipliers are not directly comparable. - The Artificial Analysis result involves "no tricks of distillation and quantization" — attributed to "raw speed" from leveraging hardware from the ground up. - "We're just starting" — further optimizations landing over time. **Application claim:** Sub-second image generation (~0.3s in many cases) enables applications "otherwise not possible because of slow inference." --- ## Calls to Action (stated) - Join the open source community: **700,000+ lines** of kernel code, all model definitions, and Max Serve open-sourced. - AI skills available to port from familiar languages into Mojo/Max. - Hiring across levels (kernel engineers to Kubernetes). - "This is not demo-ware or slideware" — `pip install modular` and validate the results independently. --- ## Unresolved Tensions / Notes - The talk is a vendor presentation; every performance figure is speaker-claimed. The only stated third-party check is the **Artificial Analysis partnership**, which the speaker himself frames as a partnership — the independence of verification is asserted, not demonstrated, within the transcript. - The speaker's own caveat about cross-vendor multiplier comparisons (7× vs. 4× with differing hardware costs) applies in principle to all throughput/latency figures presented. - Triton is credited with "70% of the Perf" while Mojo's exact standing on that quadrant chart is qualitative ("sweet spot," "best of both worlds") rather than quantified. # Speaker & Provenance Record — *Taming the AI Hardware Cambrian Explosion — Abdul Dakkak, Modular* ## Speakers & Roles | Speaker | Role | Evidence | |---|---|---| | **Abdul Dakkak** | **Creator / presenter**, self-identified Modular employee | Opens with: *"hi, my name is Abdul. I'm from modular"* | - Single-speaker talk. No guests, interviewers, or narrator beyond the presenter. - Presentation context: appears to be a **conference/vendor talk** (references a "booth behind this curtain," Q&A invitation, hiring pamphlets). ## Speaker Positioning & Promotional Content > ⚠️ **This is an explicit product promotion / vendor talk** — the stated goal is *"what modular is and what value it brings to you."* Promotional elements flagged: - Product promotion for **Mojo** (programming language), **Max** (inference engine), and **Modular Cloud**. - **Recruitment pitch**: *"We're hiring… pamphlets on the booth."* - **Live demos / booth demos** referenced (Flux 2 demo, image generation demo). - Performance claims made **by the speaker about his own company's products** (not third-party neutral claims except where noted below). ## Referenced Sources & Third-Party Claims Claims about third parties, as stated in the transcript (unverified within source): - **Independent verification claim**: *"Artificial analysis actually we've been partnering with them… they've independently been verifying our results"* — note speaker discloses a **partnership**, so independence is qualified by the speaker himself. Directs viewers to artificialanalysis.com. - Benchmark comparisons cited by speaker (all vendor-stated): vs **vLLM**, **SGLang**, **TensorRT-LLM**, **llama.cpp**, **PyTorch/diffusers**, **FlashAttention**, **cuBLAS/hipBLAS**, **Triton**, **ROCm/HIP**, **CUDA**, **OpenCL**, **SambaNova**. - Claimed open-source releases: *"over 700,000 lines of kernel code,"* model definitions, Max serve; **Mojo 1.0 open-sourcing** promised *"in the next few months"* (future commitment, not yet fact at time of speaking). - Mentioned models: **Kimi K2.5, Flux 2, Gemma 4, MiniMax 3, GLM**. ## Claims Requiring Attribution Care - All performance figures (5.5x TTFT, 1.5x throughput, 7x Flux 2, 4x AMD MI355, 3x vs next competitor, subsecond image generation) originate **solely from the speaker**, except those he attributes to Artificial Analysis. - Speaker himself concedes some comparisons are **not apples-to-apples**: *"some of this is not a fair comparison… you have to factor in the cost per GPU."* ## Availability & Access Facts - Claims **`pip install modular`** is available now for public testing. - Open source community and **AI skills** (porting tools) mentioned as available. - Comment availability: **not mentioned** in the transcript. - No curated metadata supplied beyond title; speaker identity rests on title + self-identification. ## Uncertainty Flags - Transcript is a rough auto-capture (e.g., *"Modra," "coupe losses," "Kimmy Kimmy K2.5"*) — vendor/product names normalized only where unambiguous (Modular, cuBLAS, Kimi K2.5). - "Independently verified" status of benchmarks cannot be confirmed from source; partnership disclosure limits it. Uh so let's get started. Uh hi, my name is Abdul. Uh uh I'm from modular. Uh today I'm going to be talking about what modular is and what value it brings to you. So as you probably know today there is a huge crunch uh for compute and there's like no alleviation in sight. A lot of these things are booked for the next few years. But there's also a kind of as a result you have an explosion of hardware. We call it like a Cambrian explosion. You have Nvidia providing you know a few generations of GPUs, the Intels, the ARMs, the Apples, everything uh is is present. They all have some niches uh that they cover. But the problem is that you have to ask is how do you program all of these uh systems. Now if you look at the current software as it is today um it's kind of a mess as well and that's because a lot of these hardware choices bubble up from the you know from down under uh into upwards into the stack. So this is a diagram showing how uh VLM supports both Nvidia and AMD. If you add extra hardware the the the diagram becomes even more difficult. Now today folks actually uh deploy not only using VLM but also SG lang TRTLM uh Llama CPP etc. So at Modra we thought well this is kind of a huge mess. Let's rearchitect the uh the stack from the ground up and let's build a unifying layer uh that unifies all the hardware so everything runs on our stack. To realize this, you have to actually go back to first principles and build a complete software stack with portability, performance, and programmability in mind. So at the bottommost level, we've developed Mojo, a heterogeneous programming language that's targeted towards AI as well as general purpose programming uh as well. It's kind of pythonic in style. So folks that actually know Python today pretty much know 90% of Mojo. Uh so we've built that language and that targets both CPUs, GPUs as well as all our accelerators. On top of that, we leverage Mojo to build max our inference engine. Uh so that provides you with uh AI inference for both for a bunch of different modalities like image, video, speech as well as text. We've built all our kernels underneath the hood that are built on top of Mojo. Finally, today's AI inference is distributed. So you actually need a cloud solution as well. uh you need to pick which hardware to deploy your model on, how to route uh different user requests to different uh machines, how to do KV cache offloading and and and distributed inference and disagreed inference uh to be able to serve the model at the speed and the cost that that folks expect. So putting it all together, we built a modular platform that's fully programmable. It's hardware abstracted. It's it's soda perf. Uh and then it actually can be deployed both on prem as well as on our cloud as well. So it can be if you have a cloud uh uh setup, you can actually download the modular cloud and and deployed there. So let me kind of dig into each one at a time, each part of the stack. So the first is Mojo. Uh so Mojo as I mentioned is a Pythonic system programming language. It's not unlike C++, Rust or CUDA C++. The thing is what it enables is uh programmability across different accelerators. It's blazing fast. It's again it actually competes with CUDA in terms of performance. uh and you can think of it as the best way to program CPUs and GPUs within a Python-like programming language. In terms of how we think about Mojo, uh you know, you can think about there's three different quadrants, usability, performance, and portability. Rockcom and KOD and CUDA gives you massive performance, but they sacrifice on both usability and programmability. A lot of folks use agents to actually program these system nowadays. Open CL back in you know a few years ago was popular. It gives you great portability but is sacrificed on performance and usability. Triton is has a good sweet spot that gives you 70% of the Perf uh but it gives you portability and usability and Mojo is is trying to kind of fill the niche of the sweet spot where it's the best of both worlds. It gives you the abstraction to allow you to write portable kernels and and uh and usable kernels while giving you also the low-level access uh to get the maximum performance. We've actually open sourced uh practically all of our kernels so you can actually test this stuff. You can benchmark it on your own machine. Uh we have state-of-the-art kernels on Nvidia, AMD and other hardware. So to tell you what state-of-the-art is, it means that our matrix multiplication is faster than the coupe losses, the hip losses, everything out there. And it's not by 1% in some cases by 2%. We have the fastest attention uh kernel out there, faster than flash attention for uh as well as like the AI it and things like that. Uh one thing that we've announced uh a month ago or so is Mojo 1.0, know the stabilization of our APIs. Uh, and we're going to be releasing Mojo 1.0 in the next few months. And we're also open sourcing Mojo uh, in the next few months as well. Sitting on top of Mojo is Max. So, Max is trying to kind of solve a few of the uh, assumptions that were made uh, back in the day for PyTorch. PyTorch is about 10 years old. It was not really designed for the genai inference of today. As a result, you have like VLM and SG lang that stepped into the void to give you things like KV cache, page detention and and things like that. And but all of that kind of kind of complicated the entire PyTorch ecosystem. A lot of folks don't really uh use PyTorch as is. They're resorting to these layers on top. So think of a new approach and we call that the max framework. So again it's built from the ground up uh with the best technologies with the best state-of-the-art graph compilers uh the eager mode execution. It's portable across devices. It understands your mojo programs to do novel fusion uh and it gives you the maximum performance and deployability that you're otherwise not able to do in in in regular PyTorch. So compared to other stacks uh this is a single stack that allows you to serve develop the model as well as optimize everything in one cohesive uh system. it within Max we support thousands and thousands of models in fact thousands of architectures uh everything from Gemma 4 to M uh Miniax 3 to GLM uh so any model that you think about is is practically supported in uh in in uh in Max and it has the same interface that you'd expect uh from what is standard today. So we also provide a open AI endpoints that you can use uh within the stack. If you're familiar with like PyTorch programming uh the figure on the right would or the left will not be like massive departure of your mental model. Uh you can build a model as you expect. You know you can import modules. The uh programming language for developing models is Python. So as familiar to folks uh and you can compose uh things as as you as you as you um as you build them. We also support like the eager execution to enable debugging as well. Now the other thing that one has to realize is modern LLMs uh are run on across different GPUs and in many cases multiple nodes. So also max the max APIs provide you with a way to uh program multiGPU system AC as well as multi-node systems from within the API so that the model is natively supported across uh different GPUs and different nodes as well. And this can also be heterogeneous. So there's nothing to stop you from saying the prefill is on hardware X and the decode is on hardware Y. Finally, layered on top of all of this is the modular cloud. So, as I mentioned, uh today's AI inference is distributed. It runs on many many different machines. Uh it's heterogeneous. So we provide the module cloud which is a performance native you know we think of it as a glass box because you can actually tinker with all the variables uh that are available for you in both max mojo and everything in between uh within your endpoints. So it provides provides endpoint deployment as well as bring your own cloud deployments as well. We also have a module cloud lab that allows you to experiment with all these different parameters and pick the best point in the perto frontier that you have for that model or based on your workload as well. Uh so in terms of like the UI uh if you're familiar with like any uh cloud provider I mean it's it should be pretty familiar. You just go to the website, you click on which model to you want to run, you get an AP an API key and then you plug that in uh to your HTTP request and uh you can run it. You can have dedicated endpoints as well as we can serve your own custom model. Uh you know in many cases folks have fine-tuned models or models with their own custom speculative decoders. Uh we could do all of that uh for you as well. Now, why do you care about all of this, right? It's a lot of tech. Uh why do folks care? Well, it's because you can actually uh accomplish more things uh through the system, right? So, you can actually decrease cost. You can get greater flexibility because you can actually use different hardware. You're not beholden to one hardware provider. Uh you get portability. You can use multiple different models. Uh etc. So how does Perf look like? Well, we benchmarked a lot of these uh and this is like against uh state-of-the-art, you know, the best combination of VLM or SG lang \[snorts\] in a bunch of cases. These are these benchmarks are independently verified as well. So in terms of Kimmy Kimmy K2.5, we're 5.5 faster on TTFT uh compared to VLM. We can also get 1.5x throughput improvements. So the token per minutes on per GPU is 1.5x faster uh on B200 than it is on VLM. In terms of Flux 2, and I encourage folks to actually visit the booth behind this curtain, uh there's a nice demo of Flux 2. uh we're about 7x faster uh compared to ters that compile with diffusers. So again, Max supports different modalities within the same framework. So you don't have to use an omni variant of a of a piece of software or anything like that. We support both video generation, audio generation, image generation as well as text generation. Uh we also support you know things on AMD where you can get a TCO benefit as well where we're 4x faster than the equivalent uh on um on on the MI355 systems. Actually you know some of this is not a fair comparison because you know some folks might say oh 7x is much larger than 4x but this is you cannot do that comparison very easily. uh that you can you have to factor in the cost uh per uh GPU and AMD is usually 30% cheaper than uh the Nvidia equivalent. Uh artificial analysis actually we've been partnering with them uh and we've been and they've independently been verifying our results. Uh this is our results like as of uh today. Uh we're about 3x faster than the next competitor uh on the flux 2 dev model as well. And this is not playing tricks of distillation and you know quantization or anything like that. This is just raw speed because we can leverage the hardware from the ground up and make sure that uh we're getting the max uh the max PF benefit from the model and you can actually go to artificial analysis today and see and see these results. Now what does that mean to folks again? It means you can open up applications that are otherwise kind of not possible because of slow inference. So this is actually uh a kind of a live demo. It's just somebody's pasting in the prompts. Uh but the generation is is accurate. So you can generate things in subsec generate images in subsecond uh per in many cases these are.3 seconds. Again you can go to the booth. There's a nice demo where uh you can give a prompt and then it reads the video feed and then renders it as well. Uh the other one data point is Gemma 4. So Gemma 4 was a a recent model that was just introduced about a month ago, a month and a half ago. Uh again we're quite faster than other any other competition out there. Uh so uh you know many cases actually some of these are uh hardware companies. you know, Samanova is a hardware accelerator. Uh we're, you know, let's say 6x faster than that. And in general, we're much much faster than others. And by the way, we're for all these cases, we're just starting. There's more optimizations that we're landing uh as as as time goes on. So, we're never never stopping to improving these things. So, we have a bunch of uh places where you can get involved. You can join our open source community. So, a lot of our code is open source. We have over 700,000 lines of kernel code that we've open sourced. We've over open open sourced all of our model definitions, all of our max serve. Um, we have AI skills. So, if you're not familiar with Mojo or Max, you can actually use the AI skills to port uh from languages that you're familiar with into Mojo or Max. You can also join us. We're hiring. So, there's pamphlets on the booth where we're hiring for a variety of different positions. So, as I you probably know, we we hire from low-level kernel engineers to high level folks that work on Kubernetes and everything in between. You can actually try all of this. So, this is not uh demoare or slideware. You can pip install modular right now if you want to and try all of this stuff. And I welcome everybody to validate these results. So, with that, I kind of that's it. And uh but I'm open to questions if folks have it. And I'll also be on the booth uh for the next few hours as well.