AI Demos Research — the structured-intelligence platform. Every verdict on these pages opens to the execution behind it.
Graded 26 October 2026

Can Draxlr apply a user's follow-up constraints to a previous database question?

Draxlr kept the original 90-day condition when the user narrowed the question to Texas, but returned no orders even though three qualified. It added a Texas filter using the full state name rather than the two-letter abbreviation stored in the order system.

1 of 1 test case failed

Every test case in this scenario has a result.

Pass rate0%0 of 1 gradable
Coverage1 of 1test case with a result
0PassNo test case passed.
1FailMissed at least one thing it was expected to do; the reason and proof are with each test case below.
0Not gradableEvery result here could be graded.
0UntestedEvery test case in this scenario has a result.
The pass rate is a summary. The evidence is the test case below: what we sent, what we checked, what the tool returned, and the proof.

The test case

Each test case is judged on its own: Pass, Fail, Not gradable, or Untested. The scenario result above is based on this test case.

Narrow the same question to one regionFailEvidence
What we sent

Input 1: “Show me all orders from the last 90 days, as of September 2, 2026.”

Input 2: “Only the ones shipped to Texas.”

What the tool returned

Answer to the Texas follow-up No orders matched the Texas filter in the last 90 days. Try broadening the date range or using fewer filters. This covers the 90-day window ending September 2, 2026. SQL behind the Texas answer (step 2): the state matched against 'Texas' SELECT public.orders.id, public.orders.order_number, public.orders.customer_id, public.orders.status, public.orders.category, public.orders.item_summary, public.orders.total_amount, public.orders.discount_total, public.orders.shipping_cost, public.orders.currency, public.orders.payment_method, public.orders.carrier, public.orders.tracking_number, public.orders.delivery_address, public.orders.city, public.orders.state, public.orders.postal_code, public.orders.placed_at, public.orders.shipped_at, public.orders.estimated_delivery, public.orders.delivered_at, public.orders.refunded_at FROM public.orders WHERE public.orders.placed_at >= ('2026-09-02'::date - INTERVAL '90 days') AND public.orders.placed_at < ('2026-09-02'::date + INTERVAL '1 day') AND public.orders.state ILIKE 'Texas' ORDER BY public.orders.placed_at DESC

Added to the earlier query: AND public.orders.state ILIKE 'Texas'

Transcribed from Proof 1 · Screen recording at 0:03

Transcribed from Proof 1 · Screen recording at 0:21

Everything the tool returned, in order (4)
  1. Answer: 57 orders

    There are 57 orders from the last 90 days as of September 2, 2026; the view includes the first 25 fetched, newest first. This covers the 90-day window ending September 2, 2026, inclusive.

  2. SQL behind the 90-day list (step 1) · not part of the answer
    SELECT public.orders.id, public.orders.order_number, public.orders.customer_id, public.orders.status, public.orders.category, public.orders.item_summary, public.orders.total_amount, public.orders.discount_total, public.orders.shipping_cost, public.orders.currency, public.orders.payment_method, public.orders.carrier, public.orders.tracking_number, public.orders.delivery_address, public.orders.city, public.orders.state, public.orders.postal_code, public.orders.placed_at, public.orders.shipped_at, public.orders.estimated_delivery, public.orders.delivered_at, public.orders.refunded_at FROM public.orders WHERE public.orders.placed_at >= ('2026-09-02'::date - INTERVAL '90 days') AND public.orders.placed_at < ('2026-09-02'::date + INTERVAL '1 day') ORDER BY public.orders.placed_at DESC
  3. Answer to the Texas follow-up

    No orders matched the Texas filter in the last 90 days. Try broadening the date range or using fewer filters. This covers the 90-day window ending September 2, 2026.

  4. SQL behind the Texas answer (step 2): the state matched against 'Texas' · not part of the answer
    SELECT public.orders.id, public.orders.order_number, public.orders.customer_id, public.orders.status, public.orders.category, public.orders.item_summary, public.orders.total_amount, public.orders.discount_total, public.orders.shipping_cost, public.orders.currency, public.orders.payment_method, public.orders.carrier, public.orders.tracking_number, public.orders.delivery_address, public.orders.city, public.orders.state, public.orders.postal_code, public.orders.placed_at, public.orders.shipped_at, public.orders.estimated_delivery, public.orders.delivered_at, public.orders.refunded_at FROM public.orders WHERE public.orders.placed_at >= ('2026-09-02'::date - INTERVAL '90 days') AND public.orders.placed_at < ('2026-09-02'::date + INTERVAL '1 day') AND public.orders.state ILIKE 'Texas' ORDER BY public.orders.placed_at DESC
Expected vs. Found
✗Expected: The follow-up returns only the Texas orders from the first answer, with the first question's 90-day condition still applied.Found: It kept the 90-day window but matched the state against 'Texas' while stored states are two-letter codes, and returned no rows; three Texas orders fall in that window.
Supporting proof
Proof 1Screen recording
Screen recording: both questions, both answers, the empty Texas result and the SQL of each step
Key moments
Open original ↗
Why this result

The required subset contained three Texas orders. Draxlr searched for Texas spelled out against two-letter state abbreviations, including those for Texas, Ohio and Arizona, then presented an empty answer, not an execution error. The recording shows neither the stored abbreviations nor those orders.

Also observed in this test case

Suggested broadening the dates. In its Texas answer, Draxlr presented the empty result as fact and suggested broadening the date range or using fewer filters, which would not have helped.

Window one day wider. The Texas query's window also took in 4 June 2026, one day more than the 90 days asked for; no order was placed that day, so the first answer's count of 57 is unaffected.

Tested by Ayush Ghosh · evidence dated 26 October 2026

Configuration and setup

How this tool was set up for the run and what the test needed in place. Each row is a fact from the run's records; a fact the records do not hold is left out, not guessed.

Software that produced the output
Draxlr
Build or version
Not exposed by the hosted product
Plan or tier
Premium
Model
gpt-5.5-1
Surface
Web app
Set up before the run
The first question's 90-day order list, as of 2 September 2026, had to contain both orders shipped to Texas and orders shipped elsewhere, so the follow-up's Texas subset and the remainder were both non-empty.
AI chat setup
default AI chat; no context or training supplied
Tested
26 October 2026 · Ayush Ghosh

How this scenario is graded

How we decide Pass, Fail and Not gradable. The same rules apply to every tool tested on this scenario.

How results are decided

Each test case gets one result per tool: Pass, Fail or Not gradable. A test case we haven't run yet shows Untested. There are no partial results.

The rules
  • Pass: the tool did everything the test expected, and nothing it said contradicts the correct answer.
  • Fail: at least one expected behaviour clearly didn't happen; the test case's result says which one and quotes the tool.
  • Not gradable: our evidence couldn't settle the outcome (for example a record we needed is missing). It is never counted as a fail, and the test case's result says what's missing.

Where this sits in the benchmark

This page is one cell of a larger study: one tool, one scenario. Only this benchmark's frame appears here.

LevelNameScope
BenchmarkAI Database Agents →28 scenarios · 11 tools
CapabilityConversational Interaction →
ScenarioThe user narrows what they just asked →
ToolDraxlr →

History of this result

What has happened to this result since it was first published. Runs and grades are never overwritten: a retest or a re-grade publishes a new result and keeps the earlier one readable.

from the publication record
19 November 2026First publishedAI Database Agents v1

Act on this result

Nothing filed here edits the run or the grade. A challenge opens a review, and a review can produce a new run or a re-grade — which becomes the current result and leaves this one in the history.

This matches what I see

You run the same kind of test against your own setup and get the same behaviour.

Agree →
This does not match

Yours behaves differently. Tell us what you got, with a screenshot if you have one.

Disagree →
Point out an issue

Something here is wrong — a reference value, a transcription, a grade.

Report an issue →
Request a retest

On a newer build, a larger dataset, or your own setup.

Request a retest →
We have fixed this

Tell us what changed and we schedule a rerun of the failing test case. The old result stays as history.

Vendor notice →
Filed against this evidenceNothing yet. Challenges, counter-evidence and fix notices appear here with their outcome, and stay on the page after they are resolved.
Cite this result
aidemos.com/benchmarks/ai-database-agents/results/draxlr/the-user-narrows-what-they-just-asked · 0 pass · 1 fail · coverage 1/1 · graded 2026-10-26

The same record is available as structured data through the AI Demos MCP server, with the counts, the coverage and every per-test-case reason carried as fields.

Verify the proof files

These files support this result. Open a file to inspect the original evidence.

File fingerprints (SHA-256)

A fingerprint identifies the exact file used for this result.

Proof 1 · Screen recording1bc6dadd7aff156f99e6904896bc3a096d1eb21b346b817dd69af621cca837bb