Loading…
Browse by type
Explainer reads
Explainer posts on Yomu, summarised newest first.
- 1,363 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
AI readiness in telecommunications
Telecommunications companies are adopting AI for customer experience, network operations, and cost reduction, yet initiatives often stall before production because fragmented, ungoverned, semantically opaque data creates data debt. The post argues that AI readiness depends on a semantic layer unifying datasets and business definitions, governance, and catalog metadata across systems such as Oracle, Snowflake, Salesforce, ServiceNow, and Databricks. It presents Unity Catalog as the proposed foundation, using Delta Sharing, Lakeflow Connectors, and Lakehouse Federation to exchange, ingest, or query data without uniformly replicating it, while privilege-aware metadata and audit logging support compliance. Metric Views, lineage, tags, and glossaries give agents authoritative meanings for measures and terms such as revenue, ARPU, active user, and FTTH. The conclusion is that trustworthy operational AI requires a governed, unified data foundation and organizational commitment, not simply more capable models.
Outcomes we want to see from AI at Medium Engineering
Medium Engineering defines desired outcomes and anti-goals for adopting AI, emphasizing leverage, learning, technical ambition, rigorous reasoning, fluency, and maintainable code rather than raw output. Engineers are expected to use AI for repetitive refactoring, testing, unfamiliar systems, design exploration, code analysis, and safer cross-stack work while retaining accountability for clarity, correctness, coverage, and judgment. The guidance rejects PR, commit, line, token, and model-spend metrics as success measures, allowing them only as debugging signals. It also warns against blind trust, unreviewable code volume, loss of codebase understanding, security or privacy regressions, misaligned vendors, environmental hand-waving, and performative tool adoption. The stated north star is achieving outcomes that let Medium’s humans move with the responsiveness of a much larger team while staying true to what makes the platform worth building.
amazon.science
Diverse reasoning traces teach LLMs to make better decisions
Large language models typically rely on supervised fine-tuning with a single human-verified reasoning trace per example, which limits reasoning breadth and causes mode collapse when attempting parallel reasoning. Researchers presented set-supervised fine tuning (SSFT) and global forking policy optimization (GFPO) to teach models diverse problem-solving strategies. SSFT represents reasoning as a set of complete solution paths paired with global forking tokens, utilizing bipartite matching to achieve order-invariant training and specialized tokens. GFPO applies lightweight reinforcement learning to select the optimal forking token upfront, detaching gradients from generated rollouts to ensure stable optimization. On standard reasoning and coding benchmarks, the combined framework increased single-shot accuracy by 5% to 7% while improving pass@k diversity without degrading pass@1 accuracy.
Protecting your Supabase projects from npm supply chain attacks
The post explains why npm supply-chain attacks matter to Supabase projects: Edge Functions, the Supabase CLI, and packages including supabase-js, @supabase/ssr, and @supabase/server all depend on npm. Attackers can compromise maintainers, register typosquats, or exploit build pipelines; once malicious code reaches node_modules, lifecycle scripts may read environment variables, cloud metadata, tokens, kubeconfig, and SSH keys. The proposed defense is layered: upgrade to pnpm 11, increase minimumReleaseAge to several days, pin sensitive dependency versions, commit and review lockfiles, disable or allowlist install scripts, verify package names, and pin GitHub Actions to full commit SHAs. Supabase is publishing a canonical security guide, hardening its GitHub Actions, adding secret-handling warnings to API documentation, and coordinating communications; the post also recommends rotating credentials after a potentially compromised install and using scanners as supplementary defense.
What it feels like building with Stripe Projects
The post recounts a 60-minute attempt to build a GitHub traffic dashboard with Stripe Projects, motivated by GitHub’s 14-day limit on traffic history. Using Cursor’s agentic chat workflow alongside Projects, the builder assembled a Next.js application with Auth0 authentication, Supabase persistence, Prisma schema management, Vercel hosting, scheduled jobs, environment wiring, and periodic GitHub traffic synchronization. The experiment largely avoided direct coding and instead tested whether AI-mediated orchestration could produce a production-shaped stack quickly. As services and credentials appeared across providers, the main difficulty shifted from implementation to understanding ownership, deployment topology, environment management, and operational flow. The repositories synced, charts populated, and historical data appeared, while the experience also exposed a need for clearer generated architecture or infrastructure-as-code descriptions of the assembled system.
Speed-to-lead is a solved problem
The post argues that “speed-to-lead” is no longer the central sales constraint because AI Agents can start conversations with website prospects immediately, at any hour and in any language. It describes how shift scheduling, routing delays, CRM batch processing, and unavailable SDRs created response gaps, with Fin’s former SLA targets set at one hour for best-fit leads and forty-eight hours for others. An Agent can combine immediate engagement with qualification, discovery, and an initial demo moment in one continuous conversation, filtering low-fit leads and handing teams richer context. This shifts SDR work toward phone-based selling, relationship building, complex deal navigation, and multi-threaded stakeholder engagement. The post says teams should measure completion, qualification, contact capture, meetings, pipeline, and conversation satisfaction; Fin reports S2 pipeline volume up 77% year-on-year since enabling Fin for Sales in July.
Anthropic co-founder Chris Olah's remarks on Pope Leo XIV's encyclical "Magnifica humanitas"
At the presentation of Pope Leo XIV’s encyclical “Magnifica humanitas,” Anthropic co-founder Chris Olah argues that AI’s consequences cannot be governed by frontier labs alone. Because commercial, geopolitical, and personal incentives can conflict with safety, he calls for independent critics and sustained dialogue among religious communities, civil society, scholars, governments, and developers. He describes models as grown from human language on brain-modeled structures, making their character and interaction with society questions for the humanities as well as computer science. His three priorities are protecting displaced workers and sharing gains globally, imagining conditions for human flourishing, and investigating findings about model structures and internal states that he says resemble human neuroscience and emotions, while acknowledging their meaning remains uncertain.
How we contain Claude across products
Anthropic describes how it contains Claude across claude.ai, Claude Code, and Claude Cowork as model capabilities and access expand the potential blast radius of autonomous agents. It distinguishes user misuse, model misbehavior, and external attackers, then assigns defenses across the agent’s environment, the model, and external content such as MCP servers, plugins, and web results. The main containment patterns include ephemeral gVisor containers for claude.ai, while Claude Code’s sandboxing and Cowork’s virtual-machine design tailor isolation to different users and capabilities. Reported measurements show Claude Opus 4.7 at roughly 0.1% attack success on single prompt-injection attempts and Claude Code auto mode catching roughly 83% of overeager behaviors, but neither replaces hard environmental boundaries. The conclusion favors mature isolation primitives, least-necessary access, and defense in depth because probabilistic safeguards can miss failures and permitted egress can still expose data.
Harness, Scaffold, and the AI Agent Terms Worth Getting Right
AI-agent terminology has outpaced shared definitions, leaving terms such as harness and scaffold blurred across frameworks and contexts. The glossary separates the model, which produces text but has no memory or execution loop, from scaffolding, which defines instructions, tools, output parsing, and context management, and the harness, which runs model calls, routes tool use, and decides when to stop. It presents an agent as a model plus the surrounding system that lets it act repeatedly, while distinguishing tools, skills, sub-agents, and orchestrators by their roles. For training, it defines the environment, trainer, rollout, and reward, including verifiable or learned scores and sparse or dense feedback. Because usage is not universal, the glossary offers a practical mental model rather than enforcing one vocabulary.
AI as the Next Abstraction Layer: How I see engineering evolving at Thumbtack
Thumbtack is transitioning its engineering workflow from ad-hoc tool experimentation to end-to-end agentic software development. Unlike previous deterministic abstractions such as compilers or high-level languages, artificial intelligence introduces non-determinism where identical prompts can produce varying code and trade-offs. Because consumer marketplace experiments still require weeks for behavioral signals to stabilize, accelerated implementation primarily helps teams prepare cleaner experiment designs and parallelize hypotheses within experimentation bandwidth constraints. Consequently, day-to-day engineering responsibilities shift upstream from manual coding toward problem framing, system architecture, validation, and co-designing features with product managers. Teams maintain technical sharpness by rigorously reviewing generated code while accepting that long-term specialization above the abstraction layer will become standard across the organization.
What really matters when evaluating AI Agents for customer service?
Evaluating an AI Agent for customer service requires more than measuring accuracy, resolution rates, containment, or benchmark performance on curated datasets. A credible proof of concept should reproduce real support conditions, including multi-turn, vague, sensitive, multilingual, rephrased, and multi-source queries, while testing recovery, task tracking, and human handoffs. It should also assess the customer experience: natural, on-brand tone, trust, graceful uncertainty, and seamless transitions can differ even when quantitative scores match. Finally, teams should examine whether conversations can be reviewed for gaps, how quickly knowledge and behavior can be updated, and how the vendor supports ongoing improvement. The resulting POC demonstrates not only that the Agent works now, but that it can support long-term operational success after launch.
How Databricks Genie democratizes data access in financial services
Financial services organizations have built sophisticated lakehouses, streaming pipelines, model-serving infrastructure, and self-service BI, but access remains concentrated among technical teams. Business leaders still often rely on analysts because they may lack SQL skills, BI training, or analyst access, creating the “last mile” of data democratization. Databricks Genie addresses this gap through a conversational AI interface that converts plain-English questions into governed SQL queries executed against the Databricks Lakehouse without an analyst in the loop. It operates within Unity Catalog access policies, restricts users to authorized data, makes queries read-only, and logs interactions for audit purposes, while its semantic layer maps organizational terminology such as NIM, LTV, and NII to the organization’s meanings. The stated outcome is faster, auditable answers for business questions and usage data that can inform data-product priorities.
How security teams can report cyber risk to boards
Boards are seeking visibility into cyber risk, but technical reports often fail to connect security posture with business impact or financial exposure. The post explains that compliance and cyber risk leaders can use Databricks Genie to query vulnerability posture, asset criticality, threat intelligence, control data, and historical incident costs in a governed environment. It recommends probabilistic financial modeling, including Monte Carlo simulation, to run randomized attack scenarios and produce loss distributions; Value-at-Risk framing can make those results familiar to directors. This approach replaces qualitative red/amber/green reporting with expected-loss ranges, supports investment prioritization, and enables trend analysis and board-ready answers, while the suggested cadence combines quarterly strategic briefings, monthly operational reviews, and incident-triggered updates.
Solving the Identity Crisis for AI Agents
Uber describes its 2025 effort to give AI agents verifiable identities and preserve attribution as agents act for users across compositional, multi-step workflows. The problem is that conventional human and workload identities expose only generic service identities, dropping originating users and intermediate agents across hops and weakening audits, authorization, and incident response. The design extends Uber’s Zero Trust Architecture with an Agent Registry, an AI Agent Mesh, Security Token Service (STS), MCP Gateway, downstream systems, and an AI Gateway, while integrating identity and context propagation into the agent SDK. Workloads obtain cryptographically signed workload SVIDs from SPIRE, and STS issues short-lived, scoped JWTs for each hop; MCP Gateway authenticates and authorizes tool calls. Production metrics reported show P99 STS token-exchange latency consistently below 40 milliseconds, while the roadmap places dynamic access control and unified policy enforcement above the identity foundation.
You’ve built the media products, now make them personalized
Media companies may have launched streaming services, digital editions, and mobile apps, yet still face a digital product intelligence gap: product teams must answer behavioral data questions quickly enough to personalize experiences and improve engagement. Databricks Genie gives nontechnical product leaders a conversational interface that translates natural-language questions into SQL queries, visualizations, and actionable insights over governed enterprise data, without requiring code or analyst handoffs. The agent queries governed Delta Lake tables fed by streaming clickstreams, views, and session signals, combining event-level behavior, A/B tests, audience segments, and cross-platform data in conversational answers. The source says internal benchmark accuracy improved from 32% to over 90% through multi-LLM orchestration, specialized knowledge search, and parallel reasoning, and presents Genie as reliable enough for production personalization decisions.
The product analytics you already have
Software engineers frequently rely on separate product analytics tools and teams to answer user behavior questions, creating disjointed data models and delayed insights. Existing debugging telemetry within tools like Sentry—specifically spans, structured logs, and application metrics—can serve as the primary source for product metrics. Developers enrich request spans with business-level attributes, record discrete business events through wide structured logs, and track KPIs with metric counters and distributions. This unified telemetry links business outcomes directly to underlying traces, errors, releases, and source code lines, enabling immediate root-cause analysis and automated alerts. While multi-session cohort and funnel aggregation remains a current tooling gap, standard telemetry solves day-to-day adoption, performance, and revenue questions.
Governing AI agents at scale with Unity Catalog
With AI agents proliferating across organizational functions, governance must answer which agents access sensitive data and how they act. Traditional approaches either scatter logs and permissions across systems or restrict deployment, while agents require controls over autonomous tool use and runtime behavior. Unity Catalog and Unity AI Gateway address this through four pillars: delegated access, data-centric governance, cost intelligence, and open interoperability. The design passes user identity through agents with on-behalf-of tokens, evaluates MCP tool calls with Service Policies, applies inline guardrails, and records model, access, trace, and usage data in lakehouse tables. It also connects data quality, classification, spend, and outcomes, while supporting multiple frameworks and model providers through shared governance.
How telecom CFOs can make smarter network capex decisions with AI
Telecom CFOs face multi-year, multi-billion-dollar network capex decisions involving spectrum, fiber, and 5G densification amid technology uncertainty and competitive dynamics. The post argues that operators can improve allocation by unifying network quality, customer churn, ARPU, billing, and investment history in a queryable environment. Databricks Genie provides a governed natural-language interface over that enterprise data, allowing finance leaders without SQL or data engineering backgrounds to ask cross-domain questions at geographic granularity and receive answers from systems of record. Example analyses compare ARPU and churn after 5G densification with markets still on the deployment roadmap, while scenario modeling can test accelerated deployment against historical returns. The stated outcome is better evidence for strategic judgment, shifting conversations from generic industry benchmarks toward comparable investments in the operator’s own network.
How we develop pricing and packaging at Fin
Pricing and packaging at Fin is treated as a cross-functional research, forecasting, and strategic problem rather than a finishing step. When new monetization is needed, the process starts with qualitative buyer research to define a pricing model and metric; for Fin, that produced value-based pricing tied to “outcomes,” each defined as the AI Agent successfully handling a customer service query. Quantitative willingness-to-pay studies using Gabor-Granger and Van Westendorp then indicate demand across price points, but do not establish a perfect price. Modeling combines those signals with discounting, usage, segment variation, attach rates, margins, sales capacity, competitor pricing, and finance forecasts to recommend a price alongside ARR and adoption scenarios before executive sign-off. The approach is iterative and applied selectively, while expanding products and capabilities create an ongoing need to review whether the pricing system remains coherent and aligned with delivered value.
How Databricks Genie improves retail personalization
Retail personalization often stalls when CX leaders must wait for analysts to answer questions about segment behavior, loyalty performance, channel preference, or churn risk, narrowing the window for action. Databricks Genie addresses this access gap by letting business users query unified structured and unstructured enterprise data in plain English instead of SQL. Its retail differentiators include identity-resolved queries across channels and devices, lifecycle-stage awareness, campaign response and control-group data integration, and privacy controls enforced through Unity Catalog. The source says Genie reduces routine analyst requests rather than replacing data science teams, while enabling merchandisers, category managers, loyalty marketers, and CX leaders to self-serve operational questions. It cites 7-Eleven’s use of Databricks SQL, Unity Catalog, and AI/BI Genie to launch, refine, and measure personalized offers within a secure, unified platform.
Top topics
- Architecture832
- Performance814
- LLMs658
- Machine Learning620
- Open Source563
- AI Agents546
- AI465
- Developer Experience448
Weekly digest
The best engineering reads, once a week.