{
  "version": "https://jsonfeed.org/version/1.1",
  "title": "Hopper Labs Blog",
  "home_page_url": "https://hopperlabs.ai/blog",
  "feed_url": "https://hopperlabs.ai/blog/feed.json",
  "items": [
    {
      "id": "post:publishing-test-2026-10-04",
      "url": "https://hopperlabs.ai/blog/publishing-test-2026-10-04",
      "title": "Publishing test: from source to the live blog",
      "summary": "A labeled test post tracing one piece of writing through the Hopper Labs publishing path.",
      "content_text": "A small publishing test\nThis is a deliberately labeled test post from Hopper Labs. We are using our own blog to follow one piece of writing from its source through publication to the page a reader opens.\n\nThe question is simple: can we publish a post, find it in the notebook, and trace the version that reached the website?\n\nWhat we are checking\nThe same post should appear in the blog, the feeds, and its public text and revision records. Its title, date, and body should agree across those views. Opening its link directly should work on a phone as well as on a larger screen.\n\nThe publication check records the source revision and the deployed files, then compares them with what the website serves. A successful build alone does not establish that the post is live.\n\nKeeping the trail\nThis post begins at revision 1. The revision history below and the Markdown version provide a way to inspect the published writing. If the text changes, the next revision should preserve what came before.\n\nThis is a test of the blog publishing path. It makes no claim that other site migrations or product launches are complete.",
      "authors": [
        {
          "name": "Hopper Labs",
          "url": "https://hopperlabs.ai/about"
        }
      ],
      "tags": [
        "dogfooding",
        "building"
      ],
      "date_published": "2026-10-04T12:00:00Z",
      "date_modified": "2026-10-04T12:00:00Z"
    },
    {
      "id": "post:boutique-software-at-scale",
      "url": "https://hopperlabs.ai/blog/boutique-software-at-scale",
      "title": "Boutique software at scale",
      "summary": "AI may change the economics of specificity. Empathy and feedback still decide whether the result is useful.",
      "content_text": "A thesis about specificity\nMass production made many goods more available and affordable. Standardization was often the route to scale. A highly specific solution could be expensive to create and maintain, so customers accepted something closer to the average.\n\nAI may change part of that calculation for software. If creating, adapting, and explaining a tool becomes less expensive, a more specific answer may become practical for a smaller audience.\n\nThat is a thesis to test. It is not proof that every personalized product will succeed, or that the costs of reliability, support, and maintenance disappear.\n\nGenerating more is not understanding more\nA system can produce a great deal of software without understanding why anyone would choose to use it.\n\nThe hard question remains: what is the person trying to accomplish, what makes that difficult, and what would count as an improvement? Faster implementation helps only after you can answer those questions well enough to test something.\n\nThat is why dogfooding and empathy may become more valuable as production gets cheaper. They supply the judgment that generation alone does not provide.\n\nKeep the craft where it matters\nA useful product can share a reliable foundation while adapting the parts that affect the user’s task. It does not need an entirely separate codebase for every preference.\n\nFor a learning tool, the important specificity might be the feedback. For a workspace, it might be how related people, notes, and tasks retain their context. For a calculation, it might be the way the assumptions are visible and easy to change.\n\nThe craft is in selecting the detail that changes the outcome. Adding options without judgment can move work onto the user instead of helping them.\n\nThe test is an actual choice\nGive somebody the tool for a real task. Let them keep their alternatives. Observe which one they choose, where they hesitate, and what they have to repair after using it.\n\nMeasure the result and the effort it takes to get there. Include mistakes and maintenance. A compelling first demonstration is a beginning; a product people return to is stronger evidence.\n\nIf the result is not good enough for you, find out why. If you love it but everyone else struggles, find out which part of your understanding the interface requires them to supply.\n\nWhat we are building toward\nProject Dogfood supplies a feedback loop. Raise the Floor asks whether useful starting points can reach more people. Raise the Roof asks what more demanding work becomes possible. Top Gun asks whether a specific result can be better than a strong alternative.\n\nThe bet is that these aims reinforce each other. The work on this site will have to show where that bet succeeds and where it needs revision.",
      "authors": [
        {
          "name": "Eric",
          "url": "https://hopperlabs.ai/about"
        }
      ],
      "tags": [
        "ai",
        "markets",
        "building"
      ],
      "date_published": "2026-10-02T12:00:00Z",
      "date_modified": "2026-10-03T12:00:00Z"
    },
    {
      "id": "post:choose-a-contest-you-can-measure",
      "url": "https://hopperlabs.ai/blog/choose-a-contest-you-can-measure",
      "title": "Choose a contest you can measure",
      "summary": "Project Top Gun starts with the discipline to define what winning means.",
      "content_text": "There are no points for second place\nProject Top Gun is unapologetic about trying to be the best. That ambition becomes useful when it is specific enough to test.\n\n“The best AI product” is too vague to guide a decision. “A tool that helps this student improve this essay with less wasted effort” gives you a task, a person, and a result to examine.\n\nChoose a contest that matters to somebody. Name the relevant alternatives. Then do the work.\n\nWrite down what winning means\nA comparison should define the task before seeing the results. What inputs will both tools receive? What outcome matters? Who judges it? What counts as a failure?\n\nFor an essay-feedback tool, the outcome might include whether the advice is accurate, actionable, and appropriate to the student’s goal. For a calculator, correctness and the visibility of assumptions matter. For an AI workspace, the result may include successful completion, time, cost, and the ability to inspect what happened.\n\nThose are candidate measures, not scores we have already earned. A post announcing a result should publish the actual comparison and its limits.\n\nPick a strong alternative\nA weak opponent can produce a flattering chart and teach you very little. Compare against something the intended user would realistically choose.\n\nUse the same task and disclose differences in setup. If one system needed extensive preparation, that effort belongs in the account. If a human corrected the output, that correction belongs in the result.\n\nA useful comparison can reveal that your product is better at one task and worse at another. That is information about where to specialize and where to improve.\n\nKeep ambition and judgment together\nSometimes you need pride to continue when others cannot see the point. Sometimes you need humility to recognize that the criticism is right.\n\nA defined test helps separate those situations. It does not eliminate judgment: the measure itself can be wrong, and an early result can be misleading. But it gives the argument a concrete place to begin.\n\nFailure is part of the method. Record what failed, revise the design, and compare again. A victory claim should follow the evidence rather than substitute for it.\n\nFrom philosophy to a field note\nA Top Gun field note should contain the task, the baseline, the exact setup, the results, and the next decision. Its related tool pages should identify what a reader can try today.\n\nThe project begins here with that standard. It does not announce a benchmark win. The linked products are candidates for work we can examine under it.\n\nRaise the Roof supplies the ambition to do harder work. Dogfood supplies proximity to the need. Top Gun supplies the discipline to find out whether the answer is actually better.",
      "authors": [
        {
          "name": "Eric",
          "url": "https://hopperlabs.ai/about"
        }
      ],
      "tags": [
        "excellence",
        "ai",
        "learning"
      ],
      "date_published": "2026-10-02T12:00:00Z",
      "date_modified": "2026-10-03T12:00:00Z"
    },
    {
      "id": "post:connections-should-preserve-context",
      "url": "https://hopperlabs.ai/blog/connections-should-preserve-context",
      "title": "Connections should preserve context",
      "summary": "A note, a task, and a person should keep their meaning when the work moves between tools.",
      "content_text": "The relationship is part of the work\nA task often begins in a note. A meeting involves a person, not just a copied name. A project has decisions, files, and unfinished work around it.\n\nIf moving between applications strips away those relationships, the user has to rebuild the context. The applications may each be useful while the overall workflow remains unnecessarily difficult.\n\nThe aim of connected tools is to carry the relevant relationship forward. That is a design goal to demonstrate one connection at a time.\n\nOne connection we have already described\nThe earlier Contacts-to-Calendar field note records a particular development demonstration: a contact was linked to a person, the person was selected in a calendar event, and a later change to the contact’s name and email appeared in the open event.\n\nThat example matters because it distinguishes a shared relationship from another copy. It also has a boundary. The post describes a disposable development workspace, not a public connected release or proof that every application shares every change.\n\nFollow the linked field note for the original account. The related product pages remain the place to check current availability.\n\nWhat a useful walkthrough must show\nFor a Notes-to-Tasks workflow, a useful demonstration would show the note, the resulting task, the reference back to its origin, and what happens when either item changes.\n\nFor a Calendar workflow, it would show which person is referenced, which details are copied, and which details remain live. For a graph view, it would show the actual relation and why it belongs there.\n\nThese are questions for the walkthrough, not claims that all of those behaviors are already available. A product page and a public demo should make their own scope clear.\n\nConnections need explanations\nThis blog system follows the same principle. An essay can belong to a project, appear in an ordered series, discuss a product, and build on another essay. Those relationships have different meanings.\n\nA topic label helps discovery. It does not prove that two posts make the same argument. A related post should have a stated reason. A reading path should have an order. A project should have a purpose that explains why the work belongs together.\n\nThat is why the graph includes explicit relations and the pages explain the connections in ordinary language. The map and the reading experience use the same content record.\n\nThe next useful evidence\nA future update should add a specific task, the steps somebody took, the observed result, and the limits of that observation. If the design changes, the post can be revised while its update history and stable link remain available.\n\nThe result should make the work easier to follow, both for a person returning after a month and for an agent trying to understand the source of a claim.",
      "authors": [
        {
          "name": "Eric",
          "url": "https://hopperlabs.ai/about"
        }
      ],
      "tags": [
        "connections",
        "dogfooding"
      ],
      "date_published": "2026-10-02T12:00:00Z",
      "date_modified": "2026-10-02T12:00:00Z"
    },
    {
      "id": "post:finding-the-right-sized-niche",
      "url": "https://hopperlabs.ai/blog/finding-the-right-sized-niche",
      "title": "Find the right-sized niche",
      "summary": "Specialize around a real problem. Generalize when the need and your capabilities justify it.",
      "content_text": "Specific problems deserve specific answers\nSuppose you are building a tutoring platform. “A tutor for everyone” sounds like a large market. It also conceals the task.\n\nA student preparing for the GMAT does not simply want general encouragement to learn. They want help with a particular test, its skills, their weaknesses, and the constraints of their preparation. Expertise matters because the outcome is specific.\n\nYou might begin with that niche. You might later discover that a broader education business fits your capabilities and the market better. The decision should follow evidence about the need and your ability to serve it.\n\nGeneralize along the problem\nGood generalization preserves what makes the original answer valuable. Two audiences may share an underlying skill while needing different instruction, examples, pacing, and feedback.\n\nAsk what genuinely transfers. Is it the practice engine? The explanation method? The teacher’s expertise? The customer’s reason for buying? A shared interface does not prove a shared problem.\n\nA specialization can become a family of related offerings. It can also become a generic product whose broad promise hides a weaker answer for each person. The difference is a design and operating decision, not just a marketing label.\n\nPrice and quantity describe a tradeoff\nRevenue equals price times quantity. That simple relationship clarifies one part of the choice: a small audience needs sufficient value and willingness to pay; a broad audience needs a way to serve more people economically.\n\nIt does not tell you profit. Delivery costs, acquisition, retention, support, and the cost of maintaining quality also matter. Nor does it mean that an expensive niche product is automatically better, or that a cheap broad product is automatically worse.\n\nThe practical question is whether you can provide enough value to the people you can actually reach, at a cost that lets the work continue.\n\nDogfood without becoming provincial\nDirect use gives you access to the details. It does not give you a market survey. Talk with somebody who shares the problem but not all your preferences.\n\nA narrow solution can be a strong starting point because feedback is clear. Broaden deliberately when the adjacent need is real and the capability transfers. Keep the small cohort’s reasons for loving the product visible as the audience grows.\n\nThis is the connection between Dogfood and Top Gun: proximity helps you see the problem; ambition asks you to solve the chosen problem exceptionally well.\n\nKnow when to hold your ground\nBuilding requires pride when somebody without the vision dismisses the work, and humility when somebody with understanding shows you a flaw. Neither stubbornness nor deference is sufficient on its own.\n\nWrite the prediction down. Test it. Pay attention to the person using the result. That gives you something firmer than the volume or confidence of the advice.",
      "authors": [
        {
          "name": "Eric",
          "url": "https://hopperlabs.ai/about"
        }
      ],
      "tags": [
        "markets",
        "dogfooding",
        "excellence"
      ],
      "date_published": "2026-10-02T12:00:00Z",
      "date_modified": "2026-10-03T12:00:00Z"
    },
    {
      "id": "post:raising-the-floor-and-the-roof",
      "url": "https://hopperlabs.ai/blog/raising-the-floor-and-the-roof",
      "title": "Raise the floor. Then raise the roof.",
      "summary": "A useful free baseline and advanced capability are two parts of the same ambition.",
      "content_text": "Start with something that is useful alone\nRaise the Floor is the commitment to make the starting point better: useful tools, explanations, examples, and available code that help somebody do real work.\n\nA free tool should not exist only to advertise the thing behind it. A note editor should help you write. A task tool should help you keep track of work. An explanation should leave you understanding something you did not understand before.\n\nThe project hub collects the relevant work. It does not declare every tool open source. A product’s actual license, available source, and release status belong on its product and source pages.\n\nThen make harder work possible\nRaise the Roof asks what a person can do beyond that starting point. Connected information, inspectable automation, and tools that can be adapted may help someone accomplish work that otherwise takes too much effort or coordination.\n\nMore capability should bring more understanding and control. A powerful system that hides its assumptions can make a hard task look simple while leaving the person unable to judge the answer.\n\nThis is closely related to Top Gun. Raise the Roof describes a direction for capability. Top Gun names the contest and demands evidence about the result.\n\nFree and advanced are compatible aims\nA useful free baseline can introduce a way of working. Advanced capability can serve people whose needs justify more complexity, support, or cost.\n\nThe distinction should be understandable before somebody makes a commitment. What works on its own? What requires the connected system? Which code is available, and under which terms? What is still a plan?\n\nExisting open-source permissions must remain clear. Source access under a paid license is a different promise and should be described as such. Neither label substitutes for a working release.\n\nMake the ladder visible\nThe site should help a reader move from an idea to an example, from the example to a tool, and from the tool to its source or current release information.\n\nThe research pages, blog posts, product pages, and project hubs each have a part to play. A reading path can explain the philosophy in order; the graph can reveal connections across paths. The product page supplies the concrete next step.\n\nThat is the practical purpose of these projects. They connect the principles to work someone can inspect and, where available, use.\n\nJudge the floor by the person starting there\nA better baseline is not measured by how impressive the product looks to its author. Ask whether a new user can reach a useful result, understand the limits, and continue without needing help for every step.\n\nJudge the roof by a harder task and an honest comparison. The same habit runs through both: define the need, make the attempt, and publish what happened.",
      "authors": [
        {
          "name": "Eric",
          "url": "https://hopperlabs.ai/about"
        }
      ],
      "tags": [
        "open-source",
        "connections",
        "building"
      ],
      "date_published": "2026-10-02T12:00:00Z",
      "date_modified": "2026-10-03T12:00:00Z"
    },
    {
      "id": "post:your-product-must-earn-your-use",
      "url": "https://hopperlabs.ai/blog/your-product-must-earn-your-use",
      "title": "Your product must earn your use",
      "summary": "Dogfooding begins with an honest preference, not a story about an imagined customer.",
      "content_text": "Start with a problem you can feel\nBuild something you would use. Better yet, build something you do use. That is the starting point for Project Dogfood.\n\nIt is not an absolute rule. You can serve a need you do not personally have. But then you need a real relationship with people who have it: enough curiosity to listen, enough empathy to understand what hurts, and enough ingenuity to solve it. An imagined customer is an easy person to impress. Their imagined profits are harder to spend.\n\nIf you cannot dogfood the product yourself, find somebody who can. Watch what they actually do. A polite compliment is weaker evidence than a person choosing the tool again when nobody is watching.\n\nYour habits are a starting point, not the entire market\nThe other extreme is building a perfect monument to your own peculiarities. A workflow that fits you exactly may require everyone else to become you before it becomes useful.\n\nThe useful middle is integration: feel a problem, solve it, and ask which parts belong to the problem rather than to your habits. Preserve the details that make the solution good. Remove the details that merely make it familiar to you.\n\nSometimes a very opinionated answer is the right answer. A smaller community may value it intensely. That is a choice with consequences, not a mistake by definition. You still have to understand who benefits and whether that group can support the work.\n\nChoosing another tool is feedback\nA product does not deserve your use because you built it. It earns your use by helping you accomplish the task.\n\nWhen you choose a competitor, ask what you are buying with that choice. Is it speed? Trust? A missing feature? A familiar shortcut? Better writing? Lower mental effort? The answer points toward a problem you can work on.\n\nPreference is informative, but it is not infallible. Your taste can be unusually demanding or unusually narrow. Compare your reasons with the experience of another person who has the same underlying need. Keep your standards; inspect your assumptions.\n\nMake the record useful\nA Dogfood post should name the task, show the attempt, describe the result, and explain the next change. A working demo is helpful. A failure that changes the design can be just as useful.\n\nThe project pages connect these notes to the tools, their current availability, and the next piece of reading. They are a record to build on, not a claim that every product already runs the business.",
      "authors": [
        {
          "name": "Eric",
          "url": "https://hopperlabs.ai/about"
        }
      ],
      "tags": [
        "dogfooding",
        "building",
        "markets"
      ],
      "date_published": "2026-10-02T12:00:00Z",
      "date_modified": "2026-10-02T12:00:00Z"
    },
    {
      "id": "post:contacts-to-calendar",
      "url": "https://hopperlabs.ai/blog/contacts-to-calendar",
      "title": "One person, connected across apps",
      "summary": "What we observed when a contact changed while a calendar event was open.",
      "content_text": "A small, concrete demonstration\nIn a disposable local workspace, we created a contact, explicitly created and linked a person, and selected that person in a calendar event. With the event details still open, we changed the contact name and email in another tab. The event updated to show the new details.\n\nThis demonstrated a Contacts → People → Calendar connection in the development workspace. It was not a public release, and it does not establish that every other app relationship is complete.\n\nWhy the distinction matters\nA standalone contacts app and a standalone calendar can both be useful. The connected edition adds the relationship between them, so changes can follow the person rather than requiring another manual copy.\n\nThat is the kind of behavior Pro is intended to deliver. Future walkthroughs will show more connections as they are implemented and checked.",
      "authors": [
        {
          "name": "Hopper Labs",
          "url": "https://hopperlabs.ai/about"
        }
      ],
      "tags": [
        "connections",
        "dogfooding"
      ]
    },
    {
      "id": "post:free-gre-essay-practice-scoreitnow",
      "url": "https://hopperlabs.ai/blog/free-gre-essay-practice-scoreitnow",
      "title": "Free GRE® essay practice vs ScoreItNow!",
      "summary": "ETS ScoreItNow! and Hopper Labs Essay Practice compared: price, what each includes, how grading works and privacy, so you can pick the right GRE essay practice.",
      "content_text": "Two tools, two jobs\nIf you are preparing for the GRE® Analytical Writing section, you will probably come across ScoreItNow! Online Writing Practice from ETS. You may also be looking for a free way to practice more often.\n\nThese are not really competitors. ScoreItNow! gives you ETS-written topics scored by ETS's own engine. Hopper Labs Essay Practice gives you unlimited, private practice at no cost. Many people will get the most out of using both.\n\nWhat ScoreItNow! includes\nETS lists ScoreItNow! at 20 US dollars. Using topics created by ETS, you write and submit responses to two essay topics and get immediate scores and feedback. ETS says six free bonus Analytical Writing topics are also included.\n\nETS describes the service as including sample responses and suggestions for improving your essay-writing skills. It also says its e-rater scoring engine is used in low-stakes practice products, including GRE ScoreItNow!. e-rater is the same engine ETS uses alongside human raters when scoring the real test.\n\nIf you qualify for the GRE Fee Reduction Program, ETS includes ScoreItNow! free along with two POWERPREP PLUS practice tests.\n\nWhat Hopper Labs Essay Practice includes\nHopper Labs Essay Practice is free and needs no sign-up. It runs entirely in your browser, so your essay never leaves your device.\n\nIt includes a timed test, example essays across six score levels, and a revision playground for reworking your drafts. After each essay you get a rubric report across five areas: response to the prompt, reasoning, evidence, organization and clarity.\n\nThe first session downloads about 64 MB so the model can run on your device, and it works best on a laptop or desktop.\n\nBecause nothing is uploaded, you can write about anything without worrying about who reads it, and practice as often as you like without accounts, credits or limits.\n\nHow the grading differs\nThis is the most important difference, so we will be direct. ScoreItNow! uses ETS's e-rater engine. Hopper Labs Essay Practice gives a practice grade from an on-device model that we built. It is not an official GRE score, and it is not a prediction of your GRE score.\n\nThink of our practice grade as a coach's read of a draft: useful for spotting whether your reasoning, evidence or organization is holding the essay back, and for seeing whether a revision helped. For a score from the engine ETS actually uses, use ScoreItNow!.\n\nWhen you read any feedback, from either tool, focus on patterns rather than single numbers. If reasoning or evidence is flagged essay after essay, that is where your next hour of practice should go.\n\nNeither practice tool replaces the real thing. On test day, ETS evaluates your essay with trained human raters and e-rater together.\n\nWhich one to use when\nUse Hopper Labs Essay Practice for volume and revision. Because it is free and private, you can write as many timed essays as you like, rework weak paragraphs in the revision playground, and compare your drafts to example essays at different score levels.\n\nUse ScoreItNow! for checkpoints. Once your planning and timing feel steady, use your ScoreItNow! topics to see how e-rater reads your writing on ETS-written prompts.\n\nA simple rhythm works for many people. Write two or three timed practice essays a week, revise the weakest paragraph of each, and read an example essay at the score level just above your current practice grade to see what changes.\n\nWhatever you use, practice the way you will test: 30 minutes, one Issue prompt, spell check off. ETS publishes its full Issue topic pool, so you are never short of prompts.\n\nThe bottom line\nScoreItNow! is an official ETS product that costs 20 US dollars and covers a fixed set of ETS topics. Hopper Labs Essay Practice is free and unlimited, keeps your essays on your device, and gives a practice grade with a detailed rubric report. Both can play a part in a sensible plan.\n\nIf you want to start writing right now without an account, try it in Essay Practice.\n\nGRE® is a registered trademark of ETS. This article and Hopper Labs Essay Practice are not endorsed or approved by ETS.",
      "authors": [
        {
          "name": "Hopper Labs",
          "url": "https://hopperlabs.ai/about"
        }
      ],
      "tags": [
        "learning"
      ]
    },
    {
      "id": "post:good-gre-analytical-writing-score",
      "url": "https://hopperlabs.ai/blog/good-gre-analytical-writing-score",
      "title": "What is a good GRE® Analytical Writing score?",
      "summary": "What is a good GRE AWA score? ETS percentile ranks for every Analytical Writing level (2022–2025 data), average scores by field, and how programs use the score.",
      "content_text": "Start with what the numbers mean\nThe GRE® Analytical Writing score runs from 0 to 6 in half-point steps. On its own, a number like 4.0 does not tell you much. The useful comparison is how it ranks against other test takers.\n\nETS publishes that ranking as a percentile: the percentage of test takers who scored lower than a given score. The current tables are based on everyone who tested between July 1, 2022, and June 30, 2025, which ETS reports as 785,720 Analytical Writing test takers.\n\nETS percentile ranks for each score\nFrom ETS's GRE General Test Interpretive Data, Table 1C, the percentage of test takers scoring below each level is: 6.0, 99 percent; 5.5, 98; 5.0, 93; 4.5, 85; 4.0, 63; 3.5, 40; 3.0, 16; 2.5, 7; 2.0, 3; 1.5, 1; 1.0, 1.\n\nOver the same period, the mean Analytical Writing score was 3.46, with a standard deviation of 0.85. So a 3.5 sits right around the middle, a 4.0 is above average, and every half point above 4.0 moves you past a large share of test takers.\n\nThe jumps are uneven. Going from 3.5 to 4.0 moves you from the 40th to the 63rd percentile, and from 4.0 to 4.5 moves you to the 85th. Near the top, the gains in percentile shrink because few people score there.\n\nAverages by intended field\nETS also reports average scores by intended broad graduate field, for college seniors and nonenrolled graduates in the same 2022–2025 window. Rounded to one decimal, the Analytical Writing means were: Law 3.9; Social and Behavioral Sciences 3.8; Business, Education and Life Sciences 3.7 each; Engineering, Humanities and Arts, and Physical Sciences 3.4 each.\n\nThese are averages of people who planned to study in each area, not admission requirements. They are a reasonable way to see where you stand among likely peers, and nothing more.\n\nSo what counts as good?\nGood depends on the programs you are applying to. Some publish the typical scores of admitted students; if yours do, that is the best benchmark you have. If they do not, the percentile table tells you where you stand overall.\n\nETS itself cautions against rigid cutoffs. Its guide for score users says that using a minimum GRE score as the only criterion for admission or a fellowship award is not good practice, and it recommends against cut scores used as the sole criterion. We are not going to invent a threshold either.\n\nKeep measurement error in mind. ETS reports a standard error of measurement of 0.40 for individual Analytical Writing scores. Its guidance for comparing two scores is to double the standard error of score differences, 0.57, which gives about 1.1 points; smaller differences between two applicants may not reflect a real difference in ability.\n\nHow programs use the score\nPrograms see more than a number. Schools with access to the ETS Data Manager can read applicants' actual essay responses. ETS has said schools value this because it may be the only writing sample they receive that was not coached or edited by others.\n\nETS also asks readers to treat the essay as a rough first draft, written without time to revise and without spell or grammar checking. And it notes that the Analytical Writing measure and the TOEFL Writing measure assess different skills, so the two scores are not interchangeable for applicants whose first language is not English.\n\nIn practice, this means a clearly reasoned essay helps you twice: once in the score and again if a reader looks at what you wrote.\n\nIf you want a higher score\nThe Issue scoring guide separates bands mostly by the depth of reasoning. A 4 develops a position with relevant reasons; a 5 uses logically sound reasons and well-chosen examples; a 6 develops the position fully with compelling reasons. Timed practice aimed at that difference is the most direct path upward.\n\nHopper Labs Essay Practice is free, needs no sign-up, and runs in your browser, with a timed test and a rubric report across response to the prompt, reasoning, evidence, organization and clarity. It gives a practice grade, not an official or predicted GRE score. Try it in Essay Practice.\n\nGRE® is a registered trademark of ETS. This article and Hopper Labs Essay Practice are not endorsed or approved by ETS.",
      "authors": [
        {
          "name": "Hopper Labs",
          "url": "https://hopperlabs.ai/about"
        }
      ],
      "tags": [
        "learning"
      ]
    },
    {
      "id": "post:gre-argument-essay-removed",
      "url": "https://hopperlabs.ai/blog/gre-argument-essay-removed",
      "title": "The GRE® Argument essay was removed: what changed",
      "summary": "ETS removed the GRE Argument essay on September 22, 2023. Here is the full list of changes in the shorter GRE and what they mean for Analytical Writing prep.",
      "content_text": "The short answer\nYes, the Analyze an Argument task is gone. ETS announced a shorter GRE® General Test on May 31, 2023, and the new format applied to tests beginning September 22, 2023.\n\nAnalytical Writing now consists of one 30-minute Analyze an Issue task. If you are using prep materials printed before late 2023, you can skip any section on the Argument essay.\n\nEverything that changed in the shorter GRE\nETS's announcement listed three structural changes: removal of the Analyze an Argument task from Analytical Writing, fewer questions in the Quantitative and Verbal Reasoning sections, and removal of the unscored section.\n\nETS's FAQ for schools adds more detail. Testing time was cut roughly in half, from nearly four hours to just under two. The scheduled break was eliminated. Official scores arrive faster, within 8 to 10 calendar days instead of the previous 10 to 15.\n\nWithout a scheduled break, you can still take an optional break at a test center, but ETS says the clock keeps running unless you have an approved accommodation. Optional breaks are not permitted during the at-home test.\n\nThe current test structure\nAccording to ETS, the test takes about 1 hour and 58 minutes. Analytical Writing is one Issue task in 30 minutes, and it always comes first.\n\nVerbal Reasoning has two sections: 12 questions in 18 minutes, then 15 questions in 23 minutes. Quantitative Reasoning also has two: 12 questions in 21 minutes, then 15 questions in 26 minutes. The Verbal and Quantitative sections can appear in either order after the essay.\n\nBoth measures remain section-level adaptive, meaning your performance on the first section determines the difficulty of the second.\n\nETS points out one side effect of the shorter format. Because there are fewer questions and the score scales did not change, each question counts more toward your Verbal or Quantitative score. ETS says the average time available per question is about the same as before.\n\nWhat did not change\nThe score scales stayed the same: 130 to 170 for Verbal and Quantitative, and 0 to 6 in half-point steps for Analytical Writing. ETS said this was so programs could compare applicants who tested before and after September 2023.\n\nETS also told schools that the scoring processes for all three measures did not change. The difference for writing is that your Analytical Writing score is now based on one essay instead of two.\n\nIf you took the GRE before September 22, 2023, your scores are still usable. ETS told schools to keep using scores the same way for people who tested before and after the change, and said care was taken to maintain the technical properties of the score scales.\n\nOther things held steady too. The test fee stayed the same when the shorter test launched, the retake policy is still once every 21 days and up to five times in any rolling 12-month period, and scores remain reportable for five years after your test date.\n\nWhat it means for your writing prep\nBefore the change, the Argument task asked you to evaluate someone else's reasoning, spotting assumptions and missing evidence. That is no longer tested directly, so you can put all of your writing practice into building and defending your own position.\n\nThe critical-reading habits still help. Examining assumptions is exactly what you do when you test your own thesis for weak points or take on the strongest opposing view in an Issue essay.\n\nWith one essay, there is no second task to balance out an off day. Practicing full, timed Issue essays until your planning is automatic is the most direct way to reduce that risk.\n\nSome sources still circulate old Argument prompts and templates. ETS publishes the current Issue topic pool on its website, and that is the pool you will draw from.\n\nPracticing the one task that remains\nHopper Labs Essay Practice lets you practice this kind of essay with a timed test, example essays across six score levels, and a rubric report across response to the prompt, reasoning, evidence, organization and clarity. It is free, needs no sign-up, and runs entirely in your browser.\n\nIt gives a practice grade from an on-device model, not an official or predicted GRE score. Try it in Essay Practice.\n\nGRE® is a registered trademark of ETS. This article and Hopper Labs Essay Practice are not endorsed or approved by ETS.",
      "authors": [
        {
          "name": "Hopper Labs",
          "url": "https://hopperlabs.ai/about"
        }
      ],
      "tags": [
        "learning"
      ]
    },
    {
      "id": "post:gre-essay-4-to-5",
      "url": "https://hopperlabs.ai/blog/gre-essay-4-to-5",
      "title": "How to go from a 4 to a 5 on the GRE® Issue essay",
      "summary": "A 4 is competent; a 5 is thoughtful. The GRE essay differences in position, reasoning, examples and structure, plus a revision checklist and timed drills.",
      "content_text": "What actually separates a 4 from a 5\nOn the GRE® Issue essay, a 4 is not a bad score. ETS's scoring guide calls it adequate: a clear position, developed with relevant reasons or examples, adequately organized, with acceptable clarity.\n\nA 5 is described as strong. Compare the wording. The position becomes clear and well-considered. The reasons become logically sound, and the examples well-chosen rather than merely relevant. The organization becomes focused and generally well organized, connecting ideas appropriately.\n\nIn other words, the gap is less about grammar and more about judgment. A 4 answers the question. A 5 shows it has thought about the question.\n\nMake your position specific, not just clear\nMost 4-level essays take a side: I agree, I disagree. Many 5-level essays take a side with conditions. They say when the statement holds, when it does not, and why the line falls where it does.\n\nLook at the task instructions. ETS uses several instruction types, and many ask you to consider when a claim might not hold true, or to address the strongest challenges to your view. A qualified thesis is not fence-sitting. It is the most direct way to meet those instructions.\n\nTry writing your thesis as one sentence with a because and an unless. For example: this policy makes sense for public research because accountability matters, unless the oversight costs more than the risk it prevents. That sentence already gives you a structure.\n\nGo one step deeper on every reason\nThe most common pattern in a 4 is a reason followed by an example followed by a restatement. The reason is relevant, but nothing explains why it is true or how far it goes.\n\nA 5 adds the missing step. After each example, ask yourself what it proves and what it does not. Then write that down. One sentence that explains the mechanism, such as why an incentive changes behavior, is often worth more than a second example.\n\nPick examples you can actually analyze. A specific, familiar case from history, science, your field or ordinary working life lets you discuss details. A vague appeal to many people throughout history gives you nothing to reason with.\n\nDeal with the other side honestly\nSeveral ETS instruction types explicitly ask you to address counterarguments or both views. Even when they do not, a short paragraph that takes the strongest objection seriously signals a well-considered position.\n\nAvoid the straw man. State the opposing reason in a form its supporters would accept, then explain why it does not change your conclusion, or how it narrows it. That second option, narrowing, often reads as the most thoughtful move in the whole essay.\n\nOrganize so the reader never has to guess\nA 4 is adequately organized. A 5 connects ideas. The practical difference is transitions that carry logic, not just sequence. Words like first and also only list; phrases like this matters because, even so, and which is why build an argument.\n\nGive each body paragraph one job and name it in its first sentence. If you cannot say in five words what a paragraph contributes to your thesis, cut or merge it.\n\nLanguage counts, but ETS's guidance to schools says scorers focus on the analytical logic of essays more than spelling, grammar or syntax, and weigh mechanics only when errors impede clarity. Aim for sentences that are clear and varied. Do not spend your last five minutes hunting for rare vocabulary.\n\nA revision checklist for practice essays\nAfter each practice essay, check these points before you look at any grade. Does the first paragraph state a position that answers the exact task instructions? Is there at least one condition or limit on that position?\n\nDoes every body paragraph contain a reason, a specific example, and at least one sentence explaining why the example supports the reason? Did you address the strongest opposing view in a fair form?\n\nCould a reader follow the argument from the first sentences of your paragraphs alone? Is there any sentence you had to reread to understand? Fix those first; they are the ones a rater stumbles on too.\n\nThen rewrite just one paragraph, not the whole essay. Focused revision of your weakest paragraph teaches more than a full redraft.\n\nTimed practice that builds the habit\nYou get 30 minutes on test day, with no spell checker. Practice under the same conditions: a timer, spell check off, and a prompt you have not seen before. ETS publishes its full Issue topic pool, so you can draw practice prompts from it.\n\nA workable split is about five minutes to plan your thesis and paragraph jobs, about twenty to write, and a few minutes to reread for unclear sentences. Adjust it after a few essays based on where you run short.\n\nHopper Labs Essay Practice offers a timed test, example essays across six score levels, and a revision playground for reworking a paragraph after you get feedback. It gives a practice grade from an on-device model, not an official or predicted GRE score, and your essay never leaves your device. Try it in Essay Practice.\n\nGRE® is a registered trademark of ETS. This article and Hopper Labs Essay Practice are not endorsed or approved by ETS.",
      "authors": [
        {
          "name": "Hopper Labs",
          "url": "https://hopperlabs.ai/about"
        }
      ],
      "tags": [
        "learning"
      ]
    },
    {
      "id": "post:gre-issue-essay-scoring",
      "url": "https://hopperlabs.ai/blog/gre-issue-essay-scoring",
      "title": "How the GRE® Issue essay is scored (2026)",
      "summary": "How the GRE Issue essay is scored: the 0–6 half-point scale, holistic rubric, human raters plus ETS e-rater, and what separates each score band in ETS's guide.",
      "content_text": "One essay, 30 minutes, always first\nSince September 22, 2023, the Analytical Writing section of the GRE® General Test has been a single task: Analyze an Issue, with 30 minutes to write. ETS says the Analytical Writing section always comes first, before the Verbal and Quantitative sections.\n\nYou write in a basic ETS word processor. It lets you insert and delete text, cut and paste, and undo, but it has no spell checker or grammar checker. That matters for how you practice, which we come back to at the end.\n\nYour Analytical Writing score now rests on that one essay. ETS told schools that its scoring processes did not change with the shorter test; the only difference is that the score is based on one task instead of two.\n\nThe scale: 0 to 6 in half points\nThe reported Analytical Writing score runs from 0 to 6 in half-point increments, so 3.5, 4.0 and 4.5 are all possible scores. A single score is reported for the measure.\n\nTwo special cases sit at the bottom. ETS assigns a 0 to a response that is off topic, in a foreign language, copies the topic, or consists only of keystroke characters. If the response is blank, you get NS, meaning no score.\n\nWhat holistic scoring means\nETS scores the essay holistically. In its words, scores are assigned on the basis of the overall quality of an essay in response to the assigned task. Raters do not add up points for a thesis, three examples and a conclusion. They read the whole essay and decide which score description it best fits.\n\nThat is why there is no word count or structure that guarantees a score. A tightly reasoned essay with a few typos can outscore a longer one that is clean but shallow. ETS's guidance to schools says scorers are trained to focus on the analytical logic of the essay more than on spelling, grammar or syntax, and that mechanics count only to the extent that errors get in the way of meaning.\n\nETS also describes the essay as something like a rough first draft, because test takers do not have time to fully revise. Raters read it with that in mind.\n\nHuman raters, e-rater, and disagreements\nETS's current score pages say each essay is evaluated on a 6-point holistic scale, including a score from the e-rater scoring engine, a computer program ETS built to identify essay features related to writing proficiency. ETS describes the human side as GRE raters who are carefully trained to apply the Issue scoring guide.\n\nETS research explains how the two work together. A 2014 ETS research report on the GRE describes a check-score model: the first human score is compared with e-rater, and if the two are within half a point of each other, the human score stands. If they differ by more than that, a second human reads the essay. An ETS research chapter published the same year adds that if the two humans disagree by more than a point, another human score is obtained, so the essay score always rests on human judgments.\n\nETS's current public test-taker pages do not restate that threshold for the shorter test, so treat the half-point figure as ETS's published research description rather than a guaranteed operating rule. What ETS does publish for schools is how consistent the ratings are: two ratings of an essay agree about 97 percent of the time, differ by one point about 2 percent of the time, and differ by two or more points less than 1 percent of the time.\n\nEvery response is also reviewed by trained analysts using ETS essay similarity detection software. Memorized templates and borrowed text are a risk, not a shortcut.\n\nWhat each score band emphasizes\nETS publishes an Issue scoring guide with a short description for each level. Here is the gist, in our words.\n\nA 6 is outstanding: an insightful position, developed fully with compelling reasons or persuasive examples, organized so that ideas connect logically, and written fluently and precisely. Minor errors are allowed.\n\nA 5 is strong: a clear, well-considered position, developed with logically sound reasons or well-chosen examples, focused and generally well organized, and clearly written, again with room for minor errors.\n\nA 4 is adequate: a clear position with relevant reasons or examples, adequate focus and organization, and language that conveys meaning with acceptable clarity even if some errors appear.\n\nA 3 is limited: the essay shows some competence but is obviously flawed, for example vague about the task, relying largely on unsupported claims, weakly organized, or unclear in places. A 2 largely ignores the task or has serious weaknesses, and a 1 shows fundamental deficiencies such as mostly irrelevant content or pervasive errors that make it incoherent.\n\nNotice the verbs climbing from level to level: relevant becomes sound becomes compelling; adequate becomes well organized becomes logically connected. The jump between bands is mostly about depth of reasoning, not polish.\n\nPracticing against the rubric\nBecause scoring is holistic, the most useful practice is writing full timed essays and then reading them against the band descriptions. Turn off spell check when you practice, since you will not have it on test day.\n\nHopper Labs Essay Practice lets you do that for free in your browser. It gives you a practice grade and a rubric report across response to the prompt, reasoning, evidence, organization and clarity. The grade is for practice only; it is not an official or predicted GRE score. Try it in Essay Practice.\n\nGRE® is a registered trademark of ETS. This article and Hopper Labs Essay Practice are not endorsed or approved by ETS.",
      "authors": [
        {
          "name": "Hopper Labs",
          "url": "https://hopperlabs.ai/about"
        }
      ],
      "tags": [
        "learning"
      ]
    },
    {
      "id": "post:gre-issue-two-views-prompts",
      "url": "https://hopperlabs.ai/blog/gre-issue-two-views-prompts",
      "title": "The GRE® \"two views\" Issue prompts, and how to answer them",
      "summary": "How to answer GRE Issue prompts that ask which view aligns with your own: what the task requires, a planning template, and four original practice prompts.",
      "content_text": "What a two-views prompt looks like\nMost GRE® Issue prompts give you one statement to evaluate. A two-views prompt gives you two. It usually opens with some people believe one thing, and others believe another.\n\nThe task instruction is fixed. You are asked to discuss which view more closely aligns with your own position, explain your reasoning, and, in developing and supporting your position, address both of the views presented.\n\nETS lists this as one of six instruction types used with Issue topics, alongside agree-or-disagree with a statement, a recommendation, a claim, a claim and its reason, and a policy. In the Issue topic pool ETS currently publishes, 21 of the 159 topics use the two-views instruction, so it is a real possibility on test day.\n\nWhat the instruction actually demands\nThe phrase more closely aligns is your permission to be nuanced. You do not have to accept one view completely and reject the other. You do have to say which one is closer to your own position, and why.\n\nThe phrase address both of the views is a requirement, not a suggestion. An essay that argues brilliantly for view A and never engages view B has not done the task. ETS's scoring guide repeatedly ties scores to addressing the specific task directions, so skipping half of them costs you.\n\nETS also warns that some topics appear in several versions with different instructions. Read the instruction line on test day even if the statements look familiar.\n\nCommon traps\nThe first trap is the split decision. Writing that both views have merit and stopping there does not say which view is closer to yours. Middle positions are fine, but name which way you lean.\n\nThe second trap is the straw man. If you describe the other view in a weak form, a rater can see you have not really addressed it. Give its best reason, then respond.\n\nThe third trap is treating the two views as unrelated. They usually disagree about one underlying question, such as who bears a cost or what a practice is for. Name that question and your essay becomes an analysis rather than two summaries.\n\nA planning template for 30 minutes\nMinutes 1 to 5, plan. Write one line naming the underlying question the views disagree about. Write your answer: which view is closer, and one condition that would change your mind. List one strong reason and example for your view and the best reason for the other view.\n\nIntroduction: state the disagreement in your own words and your position, including the condition. Two or three sentences is enough.\n\nBody 1 and 2: develop your view with reasons and specific examples, explaining why each example supports the reason.\n\nBody 3: present the other view at its strongest, then explain what it gets right and why it does not overturn your position, or how it limits it.\n\nConclusion: restate your qualified position in a fresh sentence. Save your last few minutes to reread for sentences a reader could misunderstand.\n\nFour practice prompts (Hopper Labs originals)\nThe prompts below were written by Hopper Labs for practice, in the style of the two-views instruction. They are not ETS prompts and are not from the ETS topic pool. Use each with the standard instruction: discuss which view more closely aligns with your own position, explain your reasoning, and address both views.\n\nPractice prompt 1. Some people believe that cities should give priority to public transit over private cars when deciding how to use street space. Others believe that street space should serve whatever form of travel most residents actually choose.\n\nPractice prompt 2. Some people believe that a scientist's first duty is to share findings as quickly as possible. Others believe that a scientist's first duty is to be certain before sharing anything at all.\n\nPractice prompt 3. Some people believe that employers should judge job applicants mainly on demonstrated skills rather than on degrees. Others believe that degrees remain the fairest signal of what an applicant can do.\n\nPractice prompt 4. Some people believe that communities preserve their character best by protecting older buildings. Others believe that a community's character lives in its people, and that buildings should give way when needs change.\n\nHow to use these prompts\nWrite each one timed, in 30 minutes, with spell check off. Then check two things before anything else: did you clearly say which view is closer to yours, and did the other view get at least one full paragraph in its strongest form?\n\nYou can write these prompts in Hopper Labs Essay Practice and get a practice grade and rubric report across response to the prompt, reasoning, evidence, organization and clarity. It is free, needs no sign-up, and runs in your browser, so your essay stays on your device. The practice grade comes from an on-device model and is not an official or predicted GRE score. Try it in Essay Practice.\n\nGRE® is a registered trademark of ETS. This article and Hopper Labs Essay Practice are not endorsed or approved by ETS.",
      "authors": [
        {
          "name": "Hopper Labs",
          "url": "https://hopperlabs.ai/about"
        }
      ],
      "tags": [
        "learning"
      ]
    },
    {
      "id": "post:open-source",
      "url": "https://hopperlabs.ai/blog/open-source",
      "title": "Open Source: useful alone",
      "summary": "Complete standalone tools, with code you can read, change, and build on.",
      "content_text": "A useful starting point\nThe free edition is built around standalone applications. A calendar should be useful as a calendar; a contacts app should help you organize people. You should not need a subscription to make either one useful.\n\nOur plan is to publish source and available downloads without requiring an account. Each release will identify its license, supported platforms, and working features. Public demos are previews; they do not mean that a source release or download is ready.\n\nStandalone does not mean closed\nFree apps do not include the native Hopper connections that keep related information in sync across the Pro workspace. You can still study the source, modify it, and build your own connections under its license.\n\nThe distinction is the connected system we maintain: shared references, coordinated behavior, and tools that work together. Existing Apache-licensed code keeps its Apache permissions. Paid additions will have their own clearly stated terms.\n\nWhat to expect here\nSource links will appear on product pages when the corresponding editions are published and checked. Workbook Studio and Ledger are also candidates for free editions; their exact release scope is still being prepared.",
      "authors": [
        {
          "name": "Hopper Labs",
          "url": "https://hopperlabs.ai/about"
        }
      ],
      "tags": [
        "open-source",
        "building"
      ]
    },
    {
      "id": "post:pro-source",
      "url": "https://hopperlabs.ai/blog/pro-source",
      "title": "Pro Source: better together",
      "summary": "The connected workspace, its source, and continuing improvements—run by you.",
      "content_text": "Connections are the product\nA person should not become a different copy every time another app needs their details. Pro is being built around apps that share references and work together: contacts, people, calendars, tasks, and AI tools in one workspace.\n\nPro includes the source for its connected app components, so you can adapt them. Ultra is a separate proposed $500-per-month edition that adds the HopperBot Studio and our engine applications. Their private source is not included.\n\nYou run it\nThe proposed Pro subscription is $200 per month for connected source access, supported desktop releases, and continuing improvements. This is a plan, not an available checkout. Final release contents and license terms will be shown before purchase.\n\nYou provide your own compatible AI account or credentials. Codex is our first integration focus; its subscription, usage limits, and eligibility remain separate from Hopper Pro. We do not include unlimited model usage or a managed business system.\n\nRenew because it is worth it\nOur cancellation commitment is simple: stop renewing when the ongoing value no longer makes sense for you. Paid access continues through the period you have paid for. You retain the right to use the versions you received under their license.\n\nCancellation ends access to future Pro updates and subscriber services. It should not disable installed local software or make your data inaccessible. A membership-service outage should not prevent you from opening your workspace.\n\nBuilt in the open, released deliberately\nProject Dogfood will show what we actually use, what works, and what still needs attention. A future managed offering may follow; there is no managed plan or service-level promise available today.",
      "authors": [
        {
          "name": "Hopper Labs",
          "url": "https://hopperlabs.ai/about"
        }
      ],
      "tags": [
        "connections",
        "building"
      ]
    },
    {
      "id": "post:project-dogfood",
      "url": "https://hopperlabs.ai/blog/project-dogfood",
      "title": "Project Dogfood",
      "summary": "Building the tools we use to build the business—and showing the work.",
      "content_text": "The business is the proving ground\nProject Dogfood is our record of building AI tools and using them to build and run the business. The goal is practical: solve problems we encounter ourselves, then make the useful parts available to others.\n\nUsing a tool ourselves is evidence about a particular workflow. It is not proof that every feature works for every person. Each post should make clear what was tried, what happened, and what remains unfinished.\n\nWhat we will share\nExpect product walkthroughs, implementation decisions, experiments, and lessons from actual use. We will link posts to the relevant tools and distinguish working releases from demonstrations and plans.\n\nWhen we use another product for a job, we will say so. The point is to build useful software and explain our decisions honestly—not to pretend that every part of the business already runs on Hopper.\n\nTwo ways to build with us\nOpen Source offers standalone tools and their code. Pro Source adds the connected system and continuing paid improvements. Both are informed by the same practical work; neither requires you to accept a claim without seeing what the product can do.",
      "authors": [
        {
          "name": "Hopper Labs",
          "url": "https://hopperlabs.ai/about"
        }
      ],
      "tags": [
        "dogfooding",
        "building"
      ]
    }
  ]
}