Tokenless

Judge a Transaction by the Company It Keeps — Somya Rai & MD Sharique, EXL

EXL developed an advanced fraud detection system that augments XGBoost with Graph Neural Network (GNN) embeddings. Traditional tabular models struggle to capture fraud signals residing in transaction neighborhoods. The system's innovative design includes meticulous sampling strategies, disjoint monthly graph construction for leakage control, and a GraphSAGE architecture that aggregates mean, max, and standard deviation of neighbor features to detect subtle fraud patterns. Emphasizing explainability for regulatory compliance and analyst trust, it also features a highly optimized serving architecture achieving 3ms P99 latency, pivoting to edge prediction to reduce cache footprint from 60GB to 2-3GB. This led to a 27-33% relative lift in true positive rates and a projected $65 million incremental fraud capture.

From 15% to 90% GPU Utilization: Fix the Data Pipeline, Not the Model

Tarun Sunkaraneni from Amazon AGI shows how a multimodal SFT pipeline that spent about 85% of its time waiting on data was brought to roughly 90% GPU utilization. He walks through concurrency with asyncio and Ray actors, prefetching, zero-copy transfer via Ray's object store, and a network-card bottleneck that appears only at scale, and argues that the hardest part of training throughput is often keeping GPUs fed rather than optimizing the model.

Small Models, Big Results: Training a Finance Agent for Under $500 — Charles Dickens, Snorkel AI

Snorkel AI and UC Berkeley's Sky Computing Lab show that a 4B Qwen3 model, trained with reinforcement learning for under $500 using the rLLM framework, beats its 235B sibling on expert-validated financial question answering built from SEC 10-K filings. The talk covers the FinQA dataset construction and verification, common failure modes of frontier models, transfer to multi-table and general tool-use tasks, and ablations showing that simple data and binary rewards outperform more elaborate setups, since the real bottleneck is disciplined tool use rather than reasoning depth.

We Kept Rebuilding Our AI Agent. Here's What Survived - Fuad Ali, Arize AI

Arize's in-product agent Alex went from a 2023 prototype to a single harness-driven agent with planning, memory, and on-demand skills. AI PM Fuad Ali walks through the decisions behind each change, including moving the plan into the harness, making finish an explicit gate, treating large tool results like files, using one small memory file per user instead of vector stores or knowledge graphs, merging specialists into one agent, and using production sessions as tests.

Stop Giving Your Agent Tools. Give It a Sandbox - Mikyo King, Arize AI

Mikyo King, head of open source at Arize AI, explains code mode, an agent architecture where the model searches for tools and composes them in a sandboxed programming environment instead of loading large tool catalogs into context. He covers why it works, how Arize built it into Phoenix with Bash, Python, SQL, and UI control, its hidden costs around secrets, permissions, durability, and double sandboxing, and Harbor evals showing 43% fewer tokens and 31% lower cost.

What a $900M Bank Error Teaches Us About AI Agents - John Feminella, Two Sigma

John Feminella of Two Sigma uses a 2020 Citibank payment error to show how AI agents can be given real autonomy through five composable controls: an agent registry, risk-based review, purpose-scoped tool access, shared evals, and classifiers that send humans only the approvals that matter.

Exploits Now Beat Patches: Open Source Security in the AI Era - Patrick Smyth, Chainguard

Patrick Smyth of Chainguard examines how frontier AI has compressed the time between vulnerability disclosure and exploitation, leaving 30-day remediation cycles obsolete. The talk covers AI-assisted attacks, waves of supply chain malware, four attack surfaces spanning malware, known CVEs, zero days, and side doors like CI/CD, developer laptops, and agent skills, and practical defenses including cooldown periods, faster CVE remediation, and signing artifacts with SLSA and Sigstore.

Can Your AI Heat My House?

A wide-ranging conversation on how AI inference is pushing cloud infrastructure, data centers, and hardware supply chains against physical limits, covering the rise of alt clouds, rack density and cooling, hardware lifecycles and circularity, talent shortages, heat reuse for communities, and what engineering leaders can do about resource-aware software.

What will computer chips look like in 2045?

Intel's leadership walks through how advanced process nodes are brought to volume, from yield learning and the lessons of 10nm to the shift from planar transistors to gate-all-around, High-NA EUV, backside power, and CFET. The discussion widens to the chip-to-chip bottleneck that hybrid bonding targets, the supply limits on EUV tooling as AI compute demand grows, and the economics that make foundry scale decisive.

The Current State of Consumer AI

The latest a16z Top 100 Consumer AI Apps report uncovers a power-user dominated economy, where a small percentage of users drive the majority of AI spending, primarily on productivity and creative tools. It highlights the rapid emergence and challenges of personal AI agents like Muse and Instinct, emphasizing privacy and high serving costs. The discussion also covers the limitations of subscription models, the burgeoning role of advertising (with OpenAI's $1B run rate), the competitive landscape of leading AI labs (ChatGPT, Claude, Gemini), and the vast untapped consumer AI white space in areas like social, dating, and entertainment, suggesting a shift from "save time" to "spend time" applications for future growth.

How VS Code Went from Monthly to Weekly Releases with AI — Harald Kirschner

Harald Kirschner, principal product manager on VS Code at Microsoft, explains how a small team serving more than 50 million users moved from monthly to weekly releases by rebuilding its whole delivery system around AI agents. He covers faster shipping through agent-ready codebases, skills, and a 10x faster TypeScript Go build, quality practices like mandatory AI code review, AI triage, auto-fix PRs and staged rollouts, and faster learning through VSC-Bench evals and daily prototyping.

Why Specialized AI Could Beat The God Model

The 5 Levels of Self-Driving Production — Eric Schwartz, Traversal

Eric Schwartz, a product manager at Traversal, argues that AI-generated code has moved engineering time from building to troubleshooting, and that observability dashboards show what broke but not why. He frames root cause analysis as a causal problem, lays out five levels of self-driving production, walks through alert intelligence at PepsiCo and incident response at American Express, and offers five questions for evaluating an AI SRE.

Why AI Didn't Actually Make You Ship Faster — Gabriel Spencer-Harper, Meticulous

Gabriel Spencer-Harper, co-founder and CEO of Meticulous, explains why assertion-based tests cannot keep pace with AI-generated code and how near-exhaustive frontend verification with zero developer effort works. The talk covers record-and-replay visual diffing on every pull request, network mocking, a deterministic browser, and coverage-guided flow selection that makes code coverage meaningful.

Your Coding Agent Is 6 Months Out of Date — Jakub Hojsan, Exa

Jakub Hojsan, a forward deployed engineer at Exa, explains how search built for coding and code-review agents closes the gap left by model knowledge cutoffs. He covers why agents need explicit rules for when to search, how query-dependent highlights cut a 100,000-character page to about 500 characters without added latency, and how Exa compares with native web search on transparency, cost, and model-provider independence, with a preview of Exa Agent.