New ways Babra helps you choose schools and prepare stronger applications.
Updated often
School matching and AI credit guidance
School matching now prioritizes official website domains. Resume parsing and AI credit prompts show when to top up or subscribe, with accurate allowance wording.
Improved
School matching by official domain
Project school matching now checks the school's official website domain first, then falls back to its name. Projects with ambiguous school names can link to the right school and display more accurate logos.
Fixed
Resume parsing credit guidance
If resume parsing fails because AI credits are insufficient, the app opens the top-up prompt instead of showing only a parsing error. Users see the reason immediately and can retry after topping up.
Improved
Options for exhausted AI credits
When AI credits run out, general prompts direct non-subscribers to subscribe to Pro and members to top up so they can continue. The same top-up message no longer appears for everyone.
Improved
Accurate school positioning limits
After users exhaust their free school positioning attempts, the paywall no longer claims unlimited use. The wording now matches the fixed allowance provided each subscription period and avoids overpromising.
Start your first free school positioning directly from the empty state, without watching a 19-minute demo video first.
Improved
Start from a simpler empty state
The 19-minute demo video is gone, and the start action now sits beside the background summary. You can begin your first free school positioning directly instead of watching a long video before trying the product.
School positioning now also matches schools to the catalog by English name. The correct school profile and logo remain linked when a Chinese name differs.
Improved
English-name catalog matching
School positioning now uses English school names as an additional way to match the school catalog. When a Chinese school name differs from the catalog, the system can still link the correct school profile and logo.
Recoverable questions and safer school list changes
Clarification questions in school selection conversations now remain available after a refresh or service recovery. List-change approvals are separate from regular questions, and prompts older than 24 hours expire automatically.
New
Recoverable clarification questions
The school selection advisor shows recoverable clarification questions above the input. They support free text, single choice, multiple choice, and temporarily skipping the question. After a page refresh or service recovery, users can keep answering without losing their current conversation progress.
Improved
Separate school list approvals
School list change confirmations are separate from regular questions. The proposal dock handles authorization, atomic application, and retries after a failure. Users can tell whether they are providing information or approving a list change, which reduces the risk of accidental changes or bypassing confirmation.
Fixed
Expired questions stop blocking chats
Clarification questions expire automatically after 24 hours and stop blocking the input or chat archiving. Users can continue the conversation or archive it instead of remaining stuck on an expired question.
When onboarding or school selection background preferences fail to save, the page now explains the reason and next step instead of showing an internal error.
Fixed
Readable save failure messages
If onboarding or school selection background preferences fail to save, the page now explains why and what to do next without displaying the server's English internal error or the user's account ID.
The landing page now matches the app's corner styling and uses sentence-case copy. Shared links also show the correct brand colors.
Improved
Rounded corners match the app
The landing page now uses 6px corners for buttons and inputs, 10px for cards, and 4px for small labels instead of square corners. The styling stays consistent when users move from the landing page into the app.
Improved
Sentence-case copy is easier to read
Uppercase labels above section headings, including IN BETA and FEATURES, have been removed. English headings and links now use sentence case, such as Every step to your dream school. The page is quieter and easier to read, and the text follows the brand's copy guidelines.
Fixed
Shared links use the correct brand colors
The social sharing preview, or OG image, now uses the brand's standard colors, so shared link cards display the correct colors.
English interfaces now show English AI progress and autosave states, while CV evaluations follow the current interface language.
Fixed
AI progress messages stay in English
With an English interface, brainstorming and essay generation now show AI progress messages in English so users can understand what the AI is doing.
Fixed
Autosave states stay in English
CV and essay editors now show autosave states in English when the interface is English, so status messages use the same language as the page.
Fixed
CV evaluations follow the interface language
CV evaluations are generated in the interface language and record that language. After a language switch, an older evaluation in another language is hidden and can be run again.
School selection in your language and preferred length
School positioning, catalog search, school lists, and programs added by link stay in your language. You can set a program length preference before recommendations and add programs to your application list without an error page.
Fixed
Keep school positioning in your language
School positioning results no longer mix in program cards from another interface language. School names, program names, and details stay in English for English users and Chinese for Chinese users.
Fixed
Search the program catalog in one language
Program catalog search now returns only entries in the current interface language, so results no longer mix English and Chinese.
Fixed
Keep a program's language stable in the school list
Programs already in the school list no longer switch language when another user researches the same program in a different language. Their display language stays stable in the list.
Improved
Research linked programs in your language
When you add a program from a link, the system creates and researches it in your language. The resulting program content stays in your language even if someone else researched the same link in another language.
New
Set program length before school positioning
School positioning onboarding and the "Background and preferences" dialog now let you select one or more program lengths: one year or less, 13-24 months, or more than 24 months. If you only want a one-year program, you can tell the AI before selection and spend less time filtering out longer programs afterward.
Fixed
Add programs without an error page
In the "Select program" drawer, "Add to application list" no longer sends the whole app to an error page. You can add the program and continue without refreshing.
AI tasks now use one billing path and keep recoverable usage records when balances change or processes and providers fail.
Improved
Unify AI usage and billing
Agent runs, messages, and batch tasks now record usage through one billing and settlement kernel. Payer selection, usage totals, and charge results are consistent across these AI features.
Fixed
Let admitted AI tasks finish
An AI task that passes its balance check can finish even if the balance falls below zero while it runs. Later tasks still check the latest balance, so long-running work records its actual usage without allowing new work when the balance is insufficient.
Fixed
Recover interrupted AI tasks safely
Long-running tasks now use execution leases that can be taken over after a retry, stop, failure, or process interruption. When a lease expires, the previous worker can no longer call the AI provider. The task can recover safely with less risk of duplicate calls, duplicate charges, or remaining stuck.
Fixed
Protect platform-funded research budgets
First free school selection and automatic program research verify eligibility for platform funding. The system reclaims a budget credential when a process stops before the task starts. Platform budgets reflect actual consumption, and interrupted research no longer stays stuck or charges the user's balance by mistake.
Fixed
Keep private CV paths out of billing
CV parsing after upload no longer writes private file paths to billing records. The CV proceeds through parsing and review without adding its file location to the usage ledger.
Fixed
Preserve recoverable AI usage evidence
AI model and billing tools now register an execution before calling the provider, then save complete, recoverable usage evidence after the response. Provider timeouts, connection interruptions, or ledger write failures no longer cause the system to silently omit usage already incurred or still awaiting review.
Filter and compare programs by length, then see the expected timing when school list building starts.
New
Filter programs by length
The school selection results list now has a multi-select "Program length" filter with All, 0-12 months, 13-24 months, and more than 24 months. You can quickly narrow the list to schools with programs that match your preferred duration.
Improved
Compare program lengths in the list
The display column controls now include a "Program length" column that you can sort. Programs with unknown lengths stay at the end of the sorted list, so you can compare durations without opening each program.
Improved
See the wait before list building
When you start building a school list, you now see that it usually takes 10-15 minutes. The expected wait is clear before the process begins, so you are less likely to leave partway through expecting a seconds-long web action.
Improved
Compare elapsed time with the estimate
The waiting panel now shows the expected range beside elapsed time. That reference lets you tell whether progress is normal or slower than expected.
Selected AI features now use DeepSeek V4 Flash or Claude Sonnet 5. CV updates use the existing re-upload path instead of whole-CV deletion.
Improved
Lower-cost models for research and editing
School list building, program and school research, and the essay and recommendation letter editor assistants now use DeepSeek V4 Flash. Its lower per-token price means the same usage consumes fewer AI credits.
Improved
Claude Sonnet 5 for document workflows
Essay and recommendation letter generation, CV parsing, and background profile AI features now use Claude Sonnet 5. The newer model also costs less per token than the previous version.
Fixed
CV updates now use re-upload
The CV page no longer offers an option to delete the entire CV and keeps the existing re-upload action. Users can update a CV directly without risking an accidental deletion that sends them back through initial setup.
Shared links display full previews, web search switches services automatically, and English guidance follows the interface language.
Fixed
Shared Babra links show full previews
Babra links shared on social platforms now show the complete large preview card. The preview image no longer fails to load.
Improved
Web search switches services automatically
If the proxy service is unavailable, web search switches to a backup search service and switches back after the proxy recovers. School positioning and other flows that rely on web research can continue when one search service fails.
Fixed
Reflection questions use English
When the interface is in English, the two reflection questions under each experience on the Personal Traits page are also in English. Users can read what to answer without guessing from Chinese prompts.
Fixed
Progress messages follow the interface language
Progress messages shown while Personal Traits are generated follow the interface language and update immediately if the language changes mid-generation. English users no longer see Chinese status text while they wait.
Clearer status labels and more accurate AI credit billing
Status labels are easier to read in light and dark modes. AI assistants now reserve and reconcile credits, messages appear immediately, and school list building bills actual model and search use.
Fixed
Readable labels in dark mode
Verified labels and pale cobalt status labels, banners, and prompt cards now use a dark blue tint with readable text in dark mode. Status markers and prompts are clear across school list building, submission tracking, recommendation letters, and other affected pages.
Improved
Consistent emphasis labels in light mode
In light mode, Verified and other emphasis labels now use the design system's cobalt text and a thin pale cobalt border. The consistent color treatment makes status labels easier to recognize across pages.
Improved
Credit reservations for essay and recommendation chats
Essay and recommendation letter AI assistants reserve credits before each reply and settle the charge from recorded usage afterward. A chat turn can still start when the balance is below the estimated cost, and the final balance may become negative.
Fixed
Interrupted charges reconcile automatically
The system now checks unfinished billing records when a chat is cancelled, exits unexpectedly, or the service restarts. Credit records stay consistent during interruptions and retries without being settled twice.
Fixed
Sent messages appear immediately
Essay and recommendation letter AI assistants now display each user message immediately after it is sent. Users can confirm that it was sent without refreshing the page.
Improved
Usage-based billing for school list building
School list building now bills model calls and search-tool use individually from actual usage, then reuses the same billing event on retries. Charges stay accurate after interruptions or retries and are not duplicated.
Material status is visible from the project list, project drawers can be bookmarked or shared, and essay content and status stay reliable during editing, imports, and AI suggestion updates.
Improved
See material status from the project list
Each project now shows the status of its statement of purpose, personal statement, and recommendation letters. You can spot missing materials without opening every project, then drag urgent projects to the top.
Improved
Open project materials without leaving the list
Select a project to open its materials drawer on the same page. The address bar includes the project parameter so you can bookmark or share the drawer, and selecting a material status dot jumps to the matching row.
Improved
Upload an existing essay from any status
The Upload existing action stays in the same place whether the essay is not started, in progress, or complete. Before replacing existing content, Babra asks for confirmation, so the upload stays easy to find and drafts are not overwritten by mistake.
Fixed
Keep unsaved edits when the essay changes elsewhere
If the same essay changes elsewhere while you are editing, the page keeps what you have typed but not saved and alerts you that the essay changed. Work in progress is not lost when another tab or an import updates the essay at the same time.
Fixed
Keep imported essays intact
Background generation no longer replaces an essay uploaded through import. If another upload starts while an import is running, Babra rejects it with a clear explanation instead of silently discarding the newly selected file.
Fixed
Keep AI edit suggestion status accurate
When the same AI edit suggestion is applied in two places at once, an edit that already took effect is no longer marked as invalid. The status shown in the interface now matches the result.
Dashboard counts now reflect real progress, and recommendation letter errors offer retry or return options. AI assistants can connect to Babra after consent and receive read-only access.
Fixed
Dashboard counts reflect your progress
The dashboard now shows the actual number of applications in progress and offers instead of leaving both metrics at zero. Submitting an application or receiving an offer is reflected in your overview.
Fixed
Recommendation letter errors offer next steps
If the recommendation letter form or editor fails to load, or the recommender no longer exists, the page shows an error with options to retry or go back instead of leaving an unusable blank screen.
Improved
Placeholder version history is hidden
The recommendation letter editor has temporarily removed the version history entry, so it no longer shows placeholder version data or a button that does not work.
New
Connect an AI assistant to Babra
You can connect an AI assistant to your Babra account and let it read your profile, program shortlist, and materials checklist. You can then ask about your application progress without copying details out of Babra.
New
Review access before approving
Before a connected application gets access, the consent screen shows its name and exactly what it can read. Separate Approve and Deny choices let you confirm the scope before granting access.
New
Connected access stays read-only
Connected applications have read-only access. They cannot change your data or spend your AI credits, so connecting an assistant cannot alter your application materials or balance.
Upload a CV and continue while it is parsed, then complete background and preference questions in one flow. Degree levels and full CV context improve school positioning without asking for the same details twice.
New
Upload your CV without waiting
New users can upload a PDF or DOCX CV on the first onboarding step and continue answering questions while Babra parses it in the background. Education and experience no longer need to be entered item by item, so your background takes fewer steps to complete.
New
Finish background questions in onboarding
Background and preference questions now appear one at a time during onboarding, with a single review on the final step. You can finish the guided flow and start school positioning without returning to fill in missing profile details.
Fixed
Degree levels stay distinct
Education entries now record degree level, including bachelor's, master's, and doctorate. The AI no longer mistakes a master's degree for a bachelor's degree, so background summaries and school positioning results are more accurate.
Improved
School positioning reads the full CV
School positioning now uses the full CV when generating recommendations. This includes research and publication details from the CV.
Improved
No duplicate experience overview
After you upload a CV, onboarding no longer asks for an experience overview. The field appears only when there is no CV, so the same experience details do not need to be entered twice.
Improved
Choose program count when positioning
The "How many programs do you plan to apply to?" question has been removed from onboarding and the background form. There is one less question to answer, and you choose the number of programs when you start school positioning.
A scannable experience list makes answers easier to manage, while more reliable saving and cleaner Personal Traits documents keep work intact and results easier to read.
Improved
A scannable experience list
The Personal Traits questionnaire now shows experiences in a list that starts collapsed, allows several sections to stay open, and pins progress plus "Jump to next unfinished section" at the bottom. You can choose what to write first without opening a full screen of fields.
Improved
A clearer representative experience marker
The "Standout / Proud" marker is now "This represents me best." That choice guides material selection for later essays. The question is easier to answer, and the result is closer to your own priorities.
Improved
One-click skipping without deletion
Skipping an experience now takes one click and automatically opens the next unfinished section. The delete option is gone; use Skip for experiences you do not want to write about without risking work you already organized.
Fixed
Answers stay saved during rapid changes
Answers no longer get lost when you check or skip immediately after typing, when a save request fails, or when you keep typing while the app saves answers before generation. Entered text stays intact, and generation uses the latest content.
Fixed
First clicks register and duplicate adds stop
After you finish typing, the first click on "Skip this experience" or the checkbox now works. Clicking "Add" twice no longer creates duplicate experiences, so one click is enough and no extra undeletable entry appears.
Fixed
Cleaner Personal Traits document spacing
Personal Traits documents no longer show full blank lines between points or an extra divider above the first section. The layout is tighter, and the fix also applies when you reopen documents generated before this update.
One Language setting controls the interface and AI output. Admins can reset stale school lists while keeping general essays, uploads, billing, and the background questionnaire. Sign-in and upgrade screens show the right links and language.
Improved
One setting for interface and AI language
Website language and AI output language are now one Language setting. The selection controls the interface, AI assistant, and generated content, so there is no second language switch to miss.
Fixed
Keep AI output in the interface language
School positioning and other AI features now follow the interface language. Chinese-interface users no longer receive English output caused by a separate setting that could default to English.
New
Reset a stale school list
Admins search by phone or email, preview a reset, then soft-delete candidates, not-interested items, application programs, their materials, essays, and recommenders. General essays, uploads, billing, and the background questionnaire remain. Users with old program data can rerun school positioning for complete details.
Fixed
Open terms and privacy links before sign-in
The Terms of Service and Privacy Policy links on the sign-in page now open the correct pages, so users can read both before signing in.
Fixed
Show verification errors in the interface language
Verification-code error messages now follow the interface language instead of always appearing in Chinese. Users on the English interface can understand why the code failed.
Fixed
Localize Pro plan durations
Plan durations in the Pro upgrade dialog now follow the interface language, so the Chinese interface no longer mixes in English duration text.
Admins can review and filter Credits usage without database queries. The school positioning list adds city and language-test comparisons, remembers visible columns, and filters tiers from a single toolbar.
New
Review a user's Credits usage
The admin console includes Credits usage history. Search by phone number or email to see a user's AI usage runs, account balance, cumulative Credits recharged, granted, and spent, plus the selected period's spending total and run count. Admins can investigate Credits usage questions without writing database SQL.
New
Filter Credits usage history
Credits usage history supports 7, 30, and 90-day presets, all time, or a custom date range, with 20 runs per page. The default view matches the billing page by showing settled runs that charged credits. A checkbox can also include failed, canceled, and unsettled runs when an investigation needs the full history.
New
Compare program cities in the list
The school positioning list has a City column. It is hidden by default, but the setting stays saved on the device after you enable it under Display columns. You can compare program cities without opening each program's details.
Improved
Compare language scores by test
The Language requirement column follows the test chosen under Display columns, such as IELTS or TOEFL. The header names the test, and each row shows its required score, including when a program does not accept it or has no minimum. This replaces truncated multi-test text with a direct comparison for the test you take.
New
Choose which school columns to show
The school positioning toolbar has a Display columns control on the right. Four information columns can be shown or hidden individually, and the device remembers your choice. Only Deadline and Tuition show by default, so the table fits on a laptop; more columns make it scroll horizontally inside the page.
Improved
Filter school tiers from one toolbar
The toolbar fits on one line. Reach, Match, and Safety are selected by default in a multi-select menu; clearing a tier hides that group. The tier filter sits beside Deadline and Display columns, with guidance on the right, so the controls no longer take a second row.
Pro Pass plans are easier to compare and try, the feedback dialog works across laptop and mobile screens, and existing SOP or PS files can go straight into the editor.
Improved
Clearer Pro Pass plan choices
The school positioning Pro Pass dialog uses the new design system and three single-select cards, with the recommended plan selected by default. One upgrade button follows your choice. Each card lists the AI school positioning features and credits it unlocks, so you can compare plans and reach payment in one step.
New
Try one program before upgrading
A fixed "Try first, decide later" option now appears at the bottom of the Pro Pass dialog. It lets you add one program manually from its official website link, so you can try the full product before deciding whether to upgrade.
Improved
Manual program additions start at 20 credits
Adding a program manually from its official website link now requires at least 20 credits instead of 100. You can try the manual add flow even with a low credit balance.
Improved
Feedback dialog matches the product
The feedback dialog now matches the rest of the product. It opens centered on desktop and as a bottom sheet on mobile, with a smaller WeChat QR code as supporting information. This gives the feedback form more prominence.
Fixed
Submit stays visible on shorter screens
The feedback dialog no longer extends past the viewport on screens 720 px tall. Its submit button is visible and clickable without scrolling inside the dialog, including on laptop screens.
New
Import an existing SOP or PS
On a writing row in Prepare Materials, you can now paste an existing SOP or PS or upload it as a PDF or DOCX. The imported document opens directly in the editor, so you can keep refining it without starting again with AI brainstorming.