shipwithjev

The shipwithjev press · 175 posts · page 4 of 5

Blog.

What the builds in the catalog add up to. Every post goes through the same gate: every number is as reported by its author.

  • Questions 31
  • Builds & people 30
  • Comparisons 24
  • Recipes 23
  • Evals & judging 12
  • Triage & routing 11
  • Classification 11
  • Guardrails 10
  • Cost 7
  • Calibration 7
  • Monitoring 6
  • Essays 3
Fig. 00 · The pressIn, judged, stamped, out
  1. FIG. 56Questions

    Does Jev Work in Languages Other Than English? What Reports Show

    Can Jev judge non-English text? What ecosystem reports and builds suggest about multilingual verdicts, plus the calibration rule that matters per language.

    Sep 30, 2026Read →

  2. FIG. 55Questions

    Does Jev Have an API? Yes: What the Ecosystem Documents

    Jev's API as the ecosystem documents it: typesafe-ai/jev via the Vercel AI Gateway, an evaluate surface, choice probabilities, and SDKs in four languages.

    Sep 30, 2026Read →

  3. FIG. 54Questions

    Can Jev Write Code? No, and Knowing Why Makes You Better at Using It

    Jev cannot write code, essays, or anything else: it's a decision model. What that means, why it's the point, and what developers use it for instead.

    Sep 30, 2026Read →

  4. FIG. 53Questions

    Can Jev Run Locally? No, and Here's the Decision Tree That Follows

    Jev is hosted-only: no local weights, no on-device inference. What that means for offline, edge, and privacy cases, and the honest alternatives for each.

    Sep 30, 2026Read →

  5. FIG. 52Questions

    Can Jev Replace My Trained Classifier? A Five-Minute Decision

    The five-minute answer to replacing a trained ML classifier with Jev: when yes, when absolutely not, and the migration that keeps both honest.

    Sep 30, 2026Read →

  6. FIG. 51Questions

    Can Jev Judge Images? Not Directly, and the Workaround Is the Lesson

    Jev judges text, not pixels. But the screenshot-free pattern, local extraction feeding text verdicts, handles more image work than you'd expect.

    Sep 30, 2026Read →

  7. FIG. 50Evals & judging

    AI Grading: Feedback at the Speed of Homework

    AI grading done honestly: rubric verdicts on short answers, instant formative feedback, and the line between grading support and grade automation.

    Sep 22, 2026Read →

  8. FIG. 49Classification

    AI Resume Screening: The Use Case That Demands Adult Supervision

    Resume screening with judge verdicts: what it does well, where bias law applies, why NYC-style audit rules exist, and the design that keeps hiring human.

    Sep 22, 2026Read →

  9. FIG. 48Cost

    Why Decisions Became Free (And What Gets Built Because of It)

    The essay version of what 600+ builds are saying: when structured judgment costs nothing, software grows a judging layer everywhere, and the map redraws.

    Sep 22, 2026Read →

  10. FIG. 47Builds & people

    Jev for Founders: The Solo Operator's Unfair Advantage Stack

    What one person can automate with a verdict machine: inbox, leads, support, content QA, and ops checks, assembled from real builds into a solo stack.

    Sep 22, 2026Read →

  11. FIG. 46Guardrails

    Jev Security and Privacy: What Leaves Your Systems, and What Shouldn't

    The security questions to ask before wiring a decision model into real data: what leaves, PII minimization, local-perception pattern, and audit posture.

    Sep 22, 2026Read →

  12. FIG. 45Builds & people

    Jev in Production: The Ops Guide the Launch Threads Skipped

    The unglamorous guide to running a decision model in production: caching, retries, versioning, monitoring, batching, and the failure modes that page you.

    Sep 22, 2026Read →

  13. FIG. 44Triage & routing

    AI Invoice Processing: The Back Office Meets the Verdict Machine

    Invoice processing as judgments: categorization, PO matching, anomaly flags, and approval routing at verdict prices, with the controls finance requires.

    Sep 22, 2026Read →

  14. FIG. 43Guardrails

    AI Spam Detection: Filtering an Adversary, Not a Category

    Spam is the one classification problem that fights back. How decision-model filters handle adversarial text, what breaks them, and the layered defense.

    Sep 22, 2026Read →

  15. FIG. 42Comparisons

    Fine-Tuning vs Prompting vs Decision Models: Picking Your Adaptation

    The three ways to make a model yours: prompting, fine-tuning, and decision-model question design. When each wins, what each costs, and the sequencing.

    Sep 22, 2026Read →

  16. FIG. 41Evals & judging

    AI Call QA Scoring: Grading Every Conversation, Not Two Percent

    Contact-center QA reviews 1-2% of calls and calls it quality. How judge verdicts on transcripts score every call for compliance, empathy, and outcomes.

    Sep 22, 2026Read →

  17. FIG. 40Triage & routing

    Churn Prediction With AI: Reading the Leaving Before the Left

    Churn prediction without the data-science project: judging cancellation language, sentiment trajectory, and risk signals in the text customers already send

    Sep 22, 2026Read →

  18. FIG. 39Classification

    Document Classification: Sorting the PDF Mountain for Cents

    Document classification with decision models: route contracts, invoices, forms, and reports by type, urgency, and risk, chunking realities included.

    Sep 22, 2026Read →

  19. FIG. 38Classification

    Entity Resolution: "Are These the Same Thing?" at a Cent a Thousand

    Entity resolution is the oldest data problem wearing new prices: record matching, dedup, and identity linking as pairwise judge verdicts, with blocking.

    Sep 22, 2026Read →

  20. FIG. 37Builds & people

    The Decision Model Glossary: Every Term This Site Uses, Defined

    The working vocabulary of the decision-model era, defined in plain language: verdicts, cascades, confidence gates, judge questions, blocking, grounding.

    Sep 22, 2026Read →

  21. FIG. 36Classification

    Intent Classification: The NLU Job That Just Got a New Engine

    Intent detection powered chatbots for a decade, badly. How decision models replace trained NLU intent classifiers: no training data, editable taxonomies.

    Sep 22, 2026Read →

  22. FIG. 35Classification

    Sentiment Analysis With LLMs: From Vibes Dashboard to Verdicts

    Sentiment analysis grew up: decision models judge emotion, intent, and churn signals per message for fractions of a cent. How to avoid the old traps.

    Sep 22, 2026Read →

  23. FIG. 34Guardrails

    Prompt Injection Detection: Judging Input Before Your AI Reads It

    Prompt injection is the SQL injection of the LLM era. How a cheap judge layer screens untrusted input before your main model reads it, and its limits.

    Sep 22, 2026Read →

  24. FIG. 33Guardrails

    LLM Guardrails: The Judge Between Your Model and the Send Button

    Guardrails are verdicts on your own AI's outputs before they reach users: policy, grounding, tone, and safety checks cheap enough to check everything.

    Sep 22, 2026Read →

  25. FIG. 32Evals & judging

    RAG Evaluation: Measuring Whether Your Retrieval Actually Grounds Anything

    RAG evaluation decomposed: retrieval relevance, answer faithfulness, and completeness as judge verdicts you can run on every query, not every quarter.

    Sep 22, 2026Read →

  26. FIG. 31Cost

    How to Reduce LLM Costs: The Playbook, Ranked by Payback

    The LLM cost-cutting playbook in payback order: cascades, right-sizing, caching, batching, and prompt diet, with receipts on what each lever saves.

    Sep 22, 2026Read →

  27. FIG. 30Builds & people

    Jev, Robotics, and the Edge: The Smallest Category With the Longest Fuse

    Six builds, one big implication: a model fast enough for game loops is flirting with control loops. The honest state of Jev in robotics and edge devices.

    Sep 22, 2026Read →

  28. FIG. 29Builds & people

    Jev Trading Bots: What People Built, and Why We Catalog It With an Eyebrow

    Within 48 hours people wired Jev to real money: a $10K live experiment, a 300ms on-chain bot. What the trading builds show, and why speed is not edge.

    Sep 22, 2026Read →

  29. FIG. 28Builds & people

    The First Week of Jev: A Launch Told Through What Got Built

    Jev launched in mid-September 2026 and the builds arrived faster than the takes. The first week's story, told through what shipped, category by category.

    Sep 22, 2026Read →

  30. FIG. 27Builds & people

    Jev for Ecommerce: Catalog, Reviews, and Support at Verdict Prices

    The ecommerce jobs that are secretly classification: product categorization, review screening, support triage, and listing moderation, priced in cents.

    Sep 22, 2026Read →

  31. FIG. 26Builds & people

    Jev for Marketers: When Growth Decisions Stop Being Vibes

    What the content-and-growth builds prove: ad teardowns at scale, pre-publish scoring, reply filtering, and marketing judgment as cheap measurable verdicts.

    Sep 22, 2026Read →

  32. FIG. 25Builds & people

    Jev for Developers: Verdicts in Your CI, PRs, and Agent Loops

    Where a decision model earns a place in a dev workflow: PR verdicts, test triage, CI quality gates, log classification, and context compaction.

    Sep 22, 2026Read →

  33. FIG. 24Essays

    Natural Language Database Queries: WHERE jev(people, 'could work from home')

    A Postgres extension puts an AI judge inside the WHERE clause: natural language filters over real rows, 129 rows a second. How it works, where it belongs.

    Sep 22, 2026Read →

  34. FIG. 23Guardrails

    AI Agent Verification: Trust, but Judge

    Agents claim they finished the task. Verification is how you know. The judge-layer pattern for checking agent work, with real builds and failure modes.

    Sep 22, 2026Read →

  35. FIG. 22Classification

    AI Data Labeling: When the Judge Becomes the Annotator

    AI data labeling flipped: models now label datasets for humans to audit. The workflow, the reported costs, and the quality controls that matter.

    Sep 22, 2026Read →

  36. FIG. 21Comparisons

    LLM vs Traditional ML for Classification: An Honest Decision Guide

    LLM or trained classifier? A 2026 guide: where each wins on cost, accuracy, and maintenance, and why decision models moved the crossover point.

    Sep 22, 2026Read →

  37. FIG. 20Recipes

    Structured Outputs From LLMs: Stop Parsing Prose for a Living

    Structured outputs from LLMs: schema modes, their failure points, and why decision models sidestep the parsing problem for classification work.

    Sep 22, 2026Read →

  38. FIG. 19Comparisons

    Jev vs Claude Haiku vs Gemini Flash: The Small-Model Bracket

    The comparison that actually matters: Jev against the fast cheap chat tiers. Interface, economics, and workload-by-workload calls, hype-free.

    Sep 22, 2026Read →

  39. FIG. 18Triage & routing

    LLM Routing: The Cascade Architecture Eating Every AI Stack

    LLM routing explained: how cheap-model-first cascades hit near-frontier accuracy at 1% of frontier cost, with confidence-gate design and real numbers.

    Sep 22, 2026Read →

  40. FIG. 17Guardrails

    AI Content Moderation: Judging Every Post Instead of Sampling

    AI content moderation used to mean sampling. At decision-model prices you judge every post, comment, and review live. Architecture, costs, and limits.

    Sep 22, 2026Read →