Loading…
Browse by type
Explainer reads
Explainer posts on Yomu, summarised newest first.
- 1,240 posts
- 20 companies
The Good Stuff
Paid picks we use ourselves. Links carry a referral code — a subscription through one earns us a commission at no extra cost to you.
Explainer posts
Building Federated Multimodal AI Workflows with NVIDIA FLARE
Federating vision-language models (VLMs) across institutions requires keeping images, text, and prompts local while coordinating training and combining updates from sites with different task and modality mixes. NVIDIA FLARE addresses this workflow through server-scheduled rounds, client-local execution, and an explicit update contract, while supporting large-object externalization, pull-based tensor streaming, and disk-backed aggregation for large payloads. The post presents FedUMM as an adapter-based design in which clients train LoRA adapters over a frozen BLIP3o backbone and send only adapter updates, while full-model training remains possible when required. In reported eight-client experiments, adapter-only federation reduced per-client communication from 28.6 GB to 0.094 GB per round, improved VQA v2 by 0.7 points, and retained about 97% of centralized performance on VQA v2 and GenEval; the evaluation used simulated sites and does not establish clinical performance or formal privacy guarantees.
Gisting: Compressing LLM Agent context to ↑ throughput and ↓ cost
Long system prompts improve agent behavior but increase inference time, cost, and the GPU capacity needed to serve a given traffic level. Shopify compresses the Sidekick GraphQL agent’s roughly 6,000-token system prompt into about 1,500 gist tokens, using special vocabulary tokens whose embeddings are learned by knowledge distillation while the model weights remain frozen. Training compares teacher logits from the full prompt with student logits from the gist-token prompt using KL divergence; at inference, the prompt is replaced with the gist-token string, requiring no custom serving path. In load tests at 350 requests per minute, compression reduced median time to first token from 438ms to 354ms and end-to-end latency from 6.8s to 4.2s, while raising throughput from 20.2 to 23.4 QPS and enabling 14% fewer GPUs in production without losing prediction quality.
The P-Completeness of Inverted Index Traversal: On the Complexity of Evaluating Boolean Query DAGs
Modern artificial intelligence agents increasingly rely on search infrastructure to run complex, neuro-symbolic reasoning workflows that compile into deeply nested, non-monotonic Boolean queries. Standard inverted index evaluation methods struggle with these query structures, facing either exponential query complexity under Document-at-a-Time models or universe-scale space penalties under Term-at-a-Time models during logical negation. To address these theoretical limitations, the author formalizes a DAG-based retrieval language named L_R and proves its query evaluation problem is strictly P-Complete. The paper introduces ComputePN, a deterministic and sparsity-aware algorithm that employs a Positive-Negative dual representation alongside native DAG memoization. This mechanism bounds evaluation time to O(|Q| · |U_active|), effectively evaluating P-Complete queries natively over the index while bypassing combinatorial tree expansion and universal scan overhead.
Run Massive-Scale UMAP in Minutes Using Multiple GPUs—Without Losing Accuracy
UMAP workflows on tens to hundreds of millions of vectors are constrained by the cost of constructing an all-neighbors k-nearest-neighbor graph, especially when repeated runs support exploration and parameter tuning. The NVIDIA cuML and cuVS 25.06 feature distributes this graph-construction stage across multiple GPUs while retaining the out-of-core design that partitions data into balanced clusters, overlaps vectors across nearby clusters, computes local graphs, and merges them into a global graph. knn_n_clusters and knn_overlap_factor expose the memory, runtime, and graph-quality trade-off: more clusters reduce per-GPU data, while more overlap can improve nearest-neighbor recall at higher computational and memory cost. On eight NVIDIA H100 GPUs, the reported evaluation found up to 74x end-to-end speedup over a projected CPU runtime for the 106M-vector MIRACL dataset, completing it in 8 minutes. The post states that performance continues to improve from one to eight GPUs while preserving comparable embedding quality across configurations.
17,600 Actions: Agent Security Is a Systems Problem
Security incidents involving AI agents, such as the OpenAI and Hugging Face event where attackers executed 17,600 actions, demonstrate that system security cannot rely entirely on manual human review. Managing agent velocity requires dedicated system-level controls to properly constrain, observe, and govern their behaviors during execution. To secure autonomous development workflows, Docker Sandboxes allow AI agents to operate within GitHub Actions, run Testcontainers tests, adjust code, and automatically generate draft pull requests inside isolated environments. In addition, escalating software supply-chain threats and the increased volume of AI-generated code drive the adoption of Docker Hardened Images and Docker Extended Lifecycle Support. These updates incorporate software built from source into container images, extend security patch coverage past end-of-life milestones, and move organizational policy enforcement directly onto developer machines.
When it comes to Governance, Retailers need a control plane for context
Retailers are moving from isolated generative-AI pilots to enterprise-wide adoption, where each store, headquarters, and digital workflow needs different data, permissions, models, and cost controls. The post argues that useful AI depends on governed business context, while fragmented tools can produce separate provider contracts, logs, permission models, cost centers, and answers to the same question. It proposes an open control plane that centralizes governance and model access without forcing employees into one application or vendor. The described Databricks design combines Unity Catalog for data and business context, Unity AI Gateway for routing and oversight, Foundation Model APIs for commercial, open-source, and custom models, and Genie for natural-language queries over governed data; the intended result is faster experimentation and deployment with visibility into access, spending, lineage, and auditability.
What's New with Monitoring in PostgreSQL 19
PostgreSQL 19, still in beta when this was written, adds monitoring and observability changes spanning lock contention, logging, WAL, wait events, remote messages, and multixacts. The default for log_lock_waits changes to on, while log_min_messages gains per-process-type levels and autoanalyze receives its own duration setting. A new wal_fpi_bytes counter exposes full-page-image volume cluster-wide, per session, per query, and in VACUUM and ANALYZE logs; new wait events expose WAL write, flush, and replay stages plus COPY file, pipe, and program I/O. Remote server messages now use normal PostgreSQL logging with origin prefixes, and pg_get_multixact_stats() surfaces multixact usage that previously required filesystem inspection and educated guessing. The post also lists upgrade work for parsers, metric collectors, wait-event dictionaries, configuration UIs, and alert thresholds, while noting that final release notes may differ.
Building Shared Memory for AI Agents in Notion
Lore is an open-source, MIT-licensed tool from Notion for giving AI agents shared, persistent memory backed by Notion, addressing experiential knowledge, follow-up tasks, decision records, and procedures that can disappear between sessions. A Lore vault uses five related Notion databases—Projects, Topics, Memories, Entities, and Facts—with MCP tools for agents, Notion or a CLI for humans, and hooks intended to retrieve context and save information in the background. The design emphasizes memory maintenance: facts expire by default, relationships such as supersedes and conflicts_with clarify updates, and the CLI can identify orphaned or duplicated information. In a 500-query SkillRet retrieval evaluation, Lore achieved 84% success, while a paired model evaluation found seeded memory recovered about 46% of failures on tasks never solved without memory. The results indicate that relevant, well-maintained memory can improve agent performance, while stale or irrelevant content can harm it.
Agent Memory Inside the Harness
The post distinguishes agent state, the run-bound working data of a single execution, from agent memory, knowledge that outlives the run and is read across sessions. Context compaction, which decides what to keep when the window fills, is lossy by construction and sets both token cost and reliability; Chroma’s study found models become unreliable well before their advertised 200k-token window, and teams compact at 95% or proactively at 60%. System prompts and skills must be versioned, governed infrastructure, not text files, with prompt registries and the SKILL.md standard for procedural memory, and skills are executable so they require sandboxing. Enterprise data enters memory through pipelines like change data capture, not migrations, with chunking, embedding, and provenance stamps. The database choice caps filtered recall under selective scoping, the consistency window, and temporal modeling, setting the ceiling for later decisions.
How Databricks Feature Store serves features with sub-second freshness
Databricks Feature Store addresses the lag between batch-computed machine-learning features and decisions that depend on events from seconds earlier. It lets a feature definition drive both offline batch processing and online streaming pipelines, orchestrating Kafka ingestion, Spark Real-Time Mode (RTM), Lakebase, and Model Serving. For a rolling 10-minute transaction sum, RTM updates per entity state in local RocksDB, expires contributions per event, and writes revised values to Lakebase through a streaming JDBC sink. The described path reports end-to-end p99 200ms latency from Kafka arrival to online feature availability, while Model Serving retrieves declared dependencies automatically at inference. Tumbling and sliding windows remain options when fewer, cheaper updates are acceptable.
Build zero-trust AI agents with Google's Agent Development Kit
Connecting an autonomous Agent Development Kit (ADK) agent to databases, internal APIs, and code execution turns model decisions into production-state changes that system prompts cannot securely constrain. The post presents an open-source Customer Support & Returns Agent built with ADK and Gemini, using an example where prompt injection could produce an unauthorized refund or leak environment secrets. Its zero-trust design combines hardware-backed signing for database mutations, gVisor isolation with no network egress and resource limits for generated Python, and deterministic semantic gateways for prompts, tool calls, and database updates. Local equivalents use HMAC keys, a database ingress verifier, a constrained Docker runner, and Python validation tests, while production mapping uses Cloud KMS, HSM, and VPC Service Controls. The conclusion is that infrastructure-enforced boundaries let the model reason dynamically within strict limits.
Measuring campaign impact with social and search data
Campaign post-mortems often report reach, engagement, impressions, and sentiment without establishing whether a campaign changed perception or behavior. It argues that social listening should measure sentiment, themes, mention volume, and audience conversation shifts, while search intelligence measures category and branded demand, promoted benefits, and share of search. A client example contrasts a campaign that generated strong social buzz while category search declined with a later campaign that reversed declining demand, captured two-thirds of share of search, and grew search intent 134%. The difference was attributed to combining socially engaging creative with media strategy and messaging designed to drive search demand. The proposed framework aligns measurement with campaign goals and uses gaps between social sentiment and search intent to identify awareness without demand, existing intent without audience expansion, or stronger market impact.
No Dumb Questions: What is AI context architecture? Why not just build your own?
AI context architecture defines the boundaries, data access, memory, and fallback behavior that constrain AI agents and make their outputs more predictable. The discussion distinguishes architecture as system design, infrastructure as the storage and delivery of context, and engineering as implementation; MCP is treated as an architectural protocol, while RAG spans all three layers. Examples include restricting retrieval to car or sports-car tires, preserving agent history, and routing incomplete or low-trust knowledge to subject-matter experts through a human-in-the-loop flow. The buy-versus-build discussion notes that coding a custom architecture is possible, but buying can provide accumulated experience with data treatment, edge cases, and recurring operational problems, while good architecture also reduces token costs and aims for consistent, predictable answers.
Insurance for Rideshare Drivers | Uber
Uber’s rideshare insurance guidance explains how coverage changes when a driver is offline, online and available, en route, or on-trip, and how personal auto insurance fits into each state. Drivers must maintain personal insurance at mandatory minimum limits, while Uber maintains commercial coverage on their behalf; repair coverage while en route or on-trip depends on comprehensive and collision coverage in the personal policy. When a driver is at fault while en route or on-trip, Uber lists at least $1 million in coverage for property damage and injuries to riders and third parties, plus vehicle-damage protection up to actual cash value with a $2,500 deductible, subject to conditions. The guidance also outlines optional protections, state-dependent coverages, certificates of insurance, and accident-reporting steps through the Driver app.
UBER FRAUDULENT USE POLICY – SEATTLE
Uber’s Seattle Fraudulent Use Policy identifies activity that may constitute fraud across rides and deliveries, including increasing trip time or distance, creating improper accounts, claiming false fees, falsifying records, and certain cancellations. Manipulating the Uber platform or GPS system with unauthorized devices, apps, or programs is also covered. Violations may cause immediate loss of access to some or all of the Marketplace Platform, holds or adjustments to suspected fraud-related earnings, restrictions on offers, or temporary deactivation during review. Affected couriers may be notified in pay statements or through in-app communication, phone, or email, and Uber may request additional information to assess a potential violation. Appeals require contacting Support or responding through the app; review typically takes up to seven business days, and the final decision is not subject to further review.
Total eclipse of the Internet: traffic impacts in Iceland, Spain, and Portugal
On August 12, a total solar eclipse crossed the North Atlantic and Europe, passing directly over Iceland, northern Spain, and Portugal. Cloudflare Radar analyzed HTTP request volumes in five-minute intervals across affected countries, comparing the data against a baseline derived from the previous three Wednesdays. Traffic reductions aligned closely with peak solar obscuration, with totality regions recording drops between 15% and 30% and specific areas shifting by up to -46.7%. Once the sun reappeared, request volumes rebounded to normal levels within minutes as users resumed device activity. The observed traffic declines were driven by aggregate human behavior rather than network failures.
What are AI Hallucinations?
AI hallucinations are coherent, confident outputs that are factually wrong, fabricated, or unsupported, appearing in chatbots, image generators, and multimodal systems. Generative models predict likely continuations rather than retrieve guaranteed facts, while training incentives, limited or flawed data, generation settings, and error compounding in reasoning models can increase the risk. Examples include Google Bard’s incorrect JWST claim, Air Canada’s nonexistent bereavement policy, Microsoft Sydney’s erratic responses, and fabricated legal citations that led to sanctions. Hallucinations create enterprise risks involving safety, liability, compliance, security, reputation, and trust; recommended controls include guardrails, source verification, adversarial testing, monitoring, human review, retrieval-augmented generation, governed data, and evaluation throughout development and deployment.
When Can LLMs Replace Humans in A/B Tests?
Replacing human participants in A/B testing with large language model predictions eliminates causal identification guarantees by design, forcing reliance on surrogate endpoint assumptions. An empirical evaluation using the Upworthy dataset revealed that raw gpt-4o-mini click-through rate predictions exhibited systematic attenuation bias, capturing only 39 percent of observed human treatment effects. To address this discrepancy, researchers formalized surrogacy and comparability conditions, finding that nonlinear machine learning calibrators like random forests and gradient-boosted trees successfully recovered treatment effects where linear ordinary least squares failed. In addition, averaging multiple sample draws per experimental unit mitigated variance and zero-bias caused by model generation temperature. Ultimately, because comparability cannot be verified for novel interventions that depart significantly from historical experiments, calibrated models cannot entirely replace human testing for genuine product innovations.
Recovery strategies to meet data residency requirements
Organizations operating under data residency requirements must balance jurisdictional compliance with disaster recovery objectives for critical workloads on AWS. In locations served by a single Region or facing multi-Region constraints, three architectural patterns offer distinct trade-offs between regulatory adherence and operational complexity. The cryptographic boundary approach replicates encrypted data to another AWS Region while restricting decryption access using AWS Key Management Service policies. The data boundary pattern deploys AWS Outposts on-premises within the source country while relying on an external parent Region control plane. Finally, strict local autonomy boundaries retain data and control planes entirely in-country through on-premises systems or multi-cloud lifeboats, requiring rigorous end-to-end continuous testing.
Why Enterprise Intelligence Needs Purpose-Built AI
Enterprise intelligence requires more than the plausible answers produced by general-purpose LLMs: it needs accurate insights from vast real-world brand data, supported by statistics, governance, and domain expertise. LLMs can hallucinate, miss context, produce inconsistent answers, and rely on limited, non-randomly sampled sources, making them unsuitable as the sole basis for high-stakes decisions. The article argues for matching technologies to tasks, using statistical models for quantitative problems such as social and online-news Reach, specialized classifiers for other tasks, and generative AI to synthesize model outputs. At enterprise scale, crisis and reputation use cases also demand near-real-time speed, consistency, predictability, explainability, auditability, monitoring, and bias testing. LLMs still have a role, including as orchestrators grounded in high-quality data, workflows, specialized models, and rigorous evaluation; the proposed future combines varied models rather than choosing between LLMs and traditional AI.
Top topics
- Architecture825
- Performance807
- LLMs636
- Machine Learning615
- Open Source555
- AI Agents522
- Developer Experience437
- AI436
Weekly digest
The best engineering reads, once a week.