Skip to content

Trust, backed by evidence.

What we published. What happened. What we learned.

Published Signals
393
Claims assessed
111
Published lessons
11
Corrections & retractions
6

What happened next

Following so far
(Hit — Dated evidence shows that every material part of the assessed forecast happened.)
0
Partly following
(Partial hit — Dated evidence shows that at least one material part of the assessed forecast happened. Other parts are still open or went differently.)
13
Against the stated outcome
(Miss — Dated evidence shows that the assessed outcome went against the forecast.)
0
Unresolved
(Unresolved — The evidence has yet to show whether the forecasted outcome happened. Plans and related developments can still be recorded as checkpoints.)
98
Public performance and coverage

93 of 393 published Signals have at least one linked claim assessment. The number of claims eligible for assessment across all Signals is unmeasured.

These totals count one current published assessment per identifiable claim, including unresolved assessments. One Signal can contain several claims. Earlier coverage, timetable changes, safety notices and editorial cases without a matched assessment have their own counts.

Latest evidence publication shown

See the same assessed claims on OTR Learns.

Site-wide scorecard

Coverage and publication checks

Publications and evidence

  • Signals

    393

    Currently published Signals from both publishing paths. Original evidence is linked in the register.

    Latest recorded

  • Early

    20

    7 dated editions. 13 separately published Workbench articles. 2 dated revisions are retained separately and are not extra articles.

    Latest recorded

  • Long Term

    40

    The source directory contains 40 articles: 36 Long Term Signals and 4 standalone Long Term articles. These publications remain part of the Signals collection; this subset is not added to its total a second time.

    Latest recorded

  • Before the Headlines

    7

    Published timing comparisons. 0 additional Workbench entries without a structured assessment in this total.

    Latest recorded

  • Came True

    2

    Published evidence comparisons. 0 additional Workbench entries without a structured assessment in this total.

    Latest recorded

  • OTR Learns

    111

    Current claim assessments shared with this scorecard. 59 curated cases and 159 published Signal notes supply context. Some cover the same claim.

    Latest recorded

  • Updated Signals

    115

    Distinct published Signals with substantive published notes. A Signal may have several notes.

    Latest recorded

  • Evidence register

    393

    Published Signals in scope. Each row links to the original. Access and dates are shown where recorded.

  • Source coverage

    Not yet measured

    Site-wide source access and completeness have no verified denominator. Source links, dates and limits are shown on individual records.

Corrections and verification

  • Corrections and retractions

    6

    Published Signal correction and retraction records. Earlier wording and recorded reasons stay in the history.

    Latest recorded

  • Research independence

    No disclosures on file

    Ownership, interests, sponsorship and conflicts are declared through the public disclosure record.

  • Public challenges

    Reader reporting

    Support takes challenges and source links. Published corrections appear in Change history. Submissions are private.

  • Workbench Signal readback

    284 of 284 checked matched

    284 of 284 current Workbench Signals have a recorded check. 0 have no recorded check for the current revision. This checks delivery of the published article.

    Latest recorded

  • Evidence publication readback

    9 of 9

    Exact publication receipts across Came True and Before the Headlines. These checks measure delivery.

    Latest recorded

  • Availability and delivery

    Not continuously measured

    Uptime, delivery to readers and independent accuracy have no verified site-wide measure here.

Reader tools and discovery
  • Home and daily popup

    Announcement published

    The popup uses the Singapore calendar day. Visitor impressions and dismissals are unmeasured.

    Latest recorded

  • Signals in partner RSS

    389

    Current Signals included in both partner feeds. Early and standalone Long Term are also distributed; they are outside this Signal-only count. Explicit withdrawals remain excluded.

    Latest recorded

  • Comtex RSS

    Same publishing stream

    The Comtex and Newstex feeds use the same current articles. Partner ingestion is confirmed separately.

  • Telegram alerts

    Opt-in alerts

    Readers can subscribe to new publications. This scorecard does not track delivery to readers.

  • Topics

    4 topics

    Public topic pages. Follow preferences are device-local.

  • Search

    Reader tool

    Search and filters are available. Search usage and result quality are unmeasured.

  • Membership

    Reader access

    Public information about reading access and membership.

  • Dashboard

    Sign-in required

    Personal reading and account activity stay private.

  • Method

    Public guide

    Read how Signals use sources and how later evidence is assessed.

  • About

    Public information

    Find OnTheRice’s publisher information and its editorial purpose.

  • Business and enquiries

    Public briefs

    The published page links to its examples and the enquiry route. Enquiries are private.

  • Public page discovery

    Public directory

    Canonical public pages are available through navigation and sitemaps. Search-engine indexing is unmeasured.

  • Privacy

    Public policy

    Read how the site handles data and what readers can control.

  • Terms

    Public policy

    Read the current terms for using OnTheRice.

From claim to learning

  1. Original claim
  2. Dated evidence
  3. Outcome
  4. Lesson
Explore the evidence

The evidence behind our work.

Follow the sources from the original claim to what we learned.

A clear trace from claim to learning

  1. Original claim

    Read what OnTheRice originally published and when.

  2. Dated evidence

    Check the source, when it was published and what it covers.

  3. Our assessment

    Compare what happened with the original claim and record its current status.

  4. What we learned

    Record what the evidence changed and what still needs checking.

Example claim record

What remains uncertain

Final public pricing, settled proceeds after costs, and Coal India’s allocation of any cash have not been verified. This is a procedural milestone, not a forecast hit.

Explore more evidence

How the system learns.

We use later evidence to improve the next assessment and record what changed.

Sample timeline

See when the Signal, its source and the assessment were published.

  1. Claim logged
  2. Evidence checked
  3. Update published
  4. Lesson recorded
  1. Signal published
  2. Source published
  3. Assessment published
  4. What we learnedRead the published lesson
Explore OTR Learns

Browse the Trust records

Evidence register

Published Signals and their evidence

393 published Signals

  1. Ten Chua Ek Kay artworks need a conservation quote. What would you price?

    First published

    4 original public sources
  2. Who handles the first refund after your app builder leaves?

    First published

    2 original public sources
  3. ProLogium corrected its funding ceiling. Cash is still unproven.

    First published

    6 original public sources
  4. Expel is testing a view of why alerts were closed. Were the decisions right?

    First published

    6 original public sources
  5. PUB wants an analyser maintained. What would count as accepted work?

    First published

    4 original public sources
  6. WHOLLY announced squeeze packs. Would they cut waste?

    First published

    7 original public sources
  7. Tronox answered the creditor story. Borrowing capacity is not cash.

    First published

    7 original public sources
  8. NEA wants autosamplers maintained. Which machines are in scope?

    First published

    8 original public sources
  9. Before a data centre buys power to bridge a grid delay, price the wait

    First published

    7 original public sources
  10. A faster freight quote can still lose money on the load

    First published

    7 original public sources
  11. Before renewing an AI supplier, ask what it keeps

    First published

    7 original public sources
  12. Therabody bought Embr’s IP. Who handles an existing Wave 2?

    First published

    13 original public sources
  13. A space data centre fund is here. Who has actually paid for computing in orbit?

    First published

    7 original public sources
  14. AliExpress added 15 sale days. Compare what shoppers keep.

    First published

    7 original public sources
  15. Replit & Co. launched with 45 creators. Membership is not a pay cheque.

    First published

    8 original public sources
  16. A cheaper AI answer needs a stopwatch, not a brochure

    First published

    8 original public sources
  17. A locked bin is not proof the paper was destroyed

    First published

    8 original public sources
  18. RMB3.26 billion of orders shipped through ZKH. What did customers pay for AI?

    First published

    11 original public sources
  19. Kadant: can replacement parts outlast acquisition-driven growth?

    First published

    9 original public sources
  20. Who gets paid when a supermarket changes its shelf labels?

    First published

    7 original public sources

Source walkthrough

What did the data-centre trial show?

The claim

Digital Realty cut total energy use by 2–3% across two 4.5 MW data halls.

Read the original Signal
The source
Singapore’s Infocomm Media Development Authority (IMDA) reported the trial result. Its case-study page is from the same publisher.
The dates
IMDA released its report on . The case-study page was updated on . OnTheRice published the Signal on .
The limits
The release supports the energy savings reported for those two halls during the trial. Long-term savings, results at other sites and demand for independent verification remain open.

What happened next

See later evidence beside what the original Signal said. Follow each case for its dates and remaining questions.

  • Equipment Purchase Agreement signed by Ainos and ASE · Partly following · Original · Later evidence

    The signed agreement records a US$9 million equipment order from ASE. It says half is due within 60 days of signing and the other half can be billed after all equipment is delivered.

    This is the same supplier order already marked partial on the original Signal, not another forecast win. The agreement does not show that ASE paid Ainos, that the order served the expansion we covered, or when the new production line first earned revenue.

Mahanadi Coalfields · early large-investor share allocation

Unresolved · Updated

Review overdue · no later published assessment recorded · Planned

What we originally said
Read final terms, compare the subsidiary’s earnings and cash economics with Coal India, and monitor announced use of proceeds.
What happened
What we originally said: Check the final sale terms and how Coal India uses any proceeds. What changed: On 18 November, Mahanadi said it had allocated about 168.8 million shares at ₹136 each to early large investors. The allocation is worth about ₹22.95 billion before costs. This is only the early allocation. It does not set the final price for the wider public offer. What that means now: Coal India is selling existing shares. This is not new funding for the mines. The allocation does not show how much cash Coal India has received or how it will use it.
What we learned
Keep the early allocation separate from the public offer, payment and the parent’s use of cash. The prospectus schedules public bidding from 19 to 23 November.
What remains open
Final public pricing, settled proceeds after costs, and Coal India’s allocation of any cash have not been verified. This is a procedural milestone, not a forecast hit.
What to check next
Look for the official final price and allotment, the seller’s settled net proceeds, and what it does with the money.
Evidence cited in the update

Read the dated update and full evidence · Read the original Signal

United States · Everee Auto Payroll

Unresolved · Updated

Review overdue · no later published assessment recorded · Planned

What we originally said
On 19 November at 22:18 SGT, Everee announced in-app Anomaly Detection, which it says flags changes such as unusual amounts or pay-rate changes before payroll approval. It did not demonstrate that a flag blocks payment or that the same flags are available in its embedded service.
What happened
What we originally said: A flag tells a reviewer where to look. We had not verified whether it stops a payment. What changed: Everee described this step when it introduced Auto Payroll in 2023. Its current product page repeats it: a flagged payment waits for a person to review it instead of going out automatically. What that means now: Keep your usual pay checks. A flag is not approval, and it is not proof that the pay is wrong.
What we learned
If you already use or are assessing this product, ask where flagged payments go in your setup and who can approve them. Keep your existing checks.
What remains open
This is Everee’s description, not a measured result. We did not verify how each connected payroll setup handles flags.
What to check next
Only an authorised demo that cannot send money can show how your setup handles a flagged draft. Use made-up details.
Evidence cited in the update

Read the dated update and full evidence · Read the original Signal

Sellforte and Exact AI · announcement dates

Unresolved · Updated

Review overdue · no later published assessment recorded · Planned

What we originally said
On 19 November 2026 at 21:44 SGT, Sellforte announced that Exact AI would lead customer onboarding in Australia, Singapore and Hong Kong. The announcement did not name a Singapore customer.
What happened
What we originally said: The partnership announcement was dated 19 November at 9:44 pm Singapore time. What changed: That was the time the newswire published the release. Sellforte’s own site dates the same partnership to 18 November and does not give a time. It mentions an Australian retailer without naming it. What that means now: The newswire time does not tell us when Sellforte first made the partnership public. Neither source shows that a Singapore customer started using it or that any ads led to extra sales.
What we learned
Ask for a fair comparison and a written test plan before paying for a new tool or changing the ad budget.
What remains open
The company page gives no hour or time zone. We found no local price or sales result.
What to check next
Look for a named Singapore customer and results from a fair comparison that shows which sales the ads added.
Evidence cited in the update
  • Sellforte’s own announcement · Primary source · Public access · Date only; same partnership announcement, not independent corroboration. · Published · Checked
  • Sellforte’s release on PR Newswire · Primary source · Public access · The timed newswire publication; no named Singapore customer. · Published · Checked

Read the dated update and full evidence · Read the original Signal

Singapore · wholesale sales

Unresolved · Updated

Review overdue · no later published assessment recorded · Planned

What we originally said
In Q2 2026 domestic wholesale sales grew 20.3% from a year earlier and foreign wholesale grew 28.6% on the same basis. Excluding petroleum the rates were 21.4% and 21.9%, respectively.
What happened
Read what happened in the published update.
What we learned
Before ordering extra stock, still check four comparable weeks of units sold in your own shop and get a written quote and restock date.
What remains open
The figures do not show demand, prices or profit at any one shop.
What to check next
Look for a change in units sold for the same item and pack, after returns. Note promotions and days the item was out of stock.
Evidence cited in the update

Read the dated update and full evidence · Read the original Signal

United States · Post refrigerated retail

Unresolved · Updated

Review overdue · no later published assessment recorded · Planned

What we originally said
The divestment changes which businesses are in the sales total, while a new product or extra stores can lift volume in the remaining operation. Only a store-level comparison can tell whether an added pack earns shelf space.
What happened
What we originally said: Post’s chilled side-dish volume rose, but that does not show whether one pack deserves more shelf space. What changed: A closer look at the same results shows that Post credited the rise in side-dish volume to new store-brand products. What that means now: Selling more after adding new packs does not show that a pack already on your shelf sold faster.
What we learned
Before giving a pack more shelf space, compare how many shoppers bought it, how much was left after costs, and how much went to waste. Compare similar weeks and keep pack sizes and promotions alike.
What remains open
No matched store trial, result for an individual pack, or retailer’s profit was verified.
What to check next
A small store trial with paid purchases, waste and costs for the same pack would tell you more than the company-wide figure.
Evidence cited in the update

Read the dated update and full evidence · Read the original Signal

Published changes and notices

22 November check of current HAIARA W5 PRO safety guidance; date the notice changed is not known.

Safety update · Updated

Scheduled review · Planned

What we originally said
The remedy requires stopping use, contacting Mitrytech, a photo with recalled written on the device, and confirmation of lawful disposal for a full refund.
What happened
What we originally said: We described the steps Mitrytech had offered for a refund. What changed: When we checked on 22 November (Singapore time), the U.S. Consumer Product Safety Commission’s recall page said Mitrytech was no longer honoring the recall. The older refund steps still appear lower on the page. Do not rely on those steps or wait for a refund before acting.
What we learned
If your black HAIARA W5 PRO has “20W” on the front and no date code on the side, stop using it. Do not pack it for a trip, sell it or give it away. Do not put it in the trash or regular recycling.
What remains open
The notice does not say when the refund instructions stopped applying. It does not establish what happened to individual refund requests.
What to check next
Call your local hazardous-waste service before taking the power bank anywhere. Ask if it accepts defective lithium-ion batteries and how to dispose of this unit safely.
Evidence cited in the update

Read the dated update and full evidence · Read the original Signal

First Hong Kong market session covered by the original check; not a forecast.

Milestone · Updated

Review overdue · no later published assessment recorded · Planned

What we originally said
Monitor the 12 November China and Hong Kong market response
What happened
Hong Kong Exchange records show that on 12 November, COSCO Shipping Development closed at HK$0.875, down about 1.13% from HK$0.885. COSCO Shipping Energy closed at HK$15.10, down about 2.96% from HK$15.56.
What we learned
This records the Hong Kong part of our original market check. It does not show that the proposed 16-tanker deal caused the moves. It also does not show shareholder approval, funding or supplier payment.
What remains open
No trade or return was verified. The exchange prices describe two listings and one day; the 16-tanker deal’s circular and approvals are separate.
What to check next
Watch the exact 16-tanker shareholder circular expected by 30 November, and keep approvals, funded orders and deliveries separate.
Evidence cited in the update

Read the dated update and full evidence · Read the original Signal

Existing participation window elapsed; no new deadline announced.

Deadline changed · Updated

Review overdue · no later published assessment recorded · Planned

What we originally said
Approach potential partners before the application deadline and seek a written scope by 4 December.
What happened
The dates in the official SNCF Réseau notice have now passed for all four lots. The notice still prints 19 November at 17:00 with a UTC+02:00 offset. Either the printed offset or the usual French winter time puts the date in the past. We did not find an official extension.
What we learned
Our original invitation to prepare before the deadline is no longer current. Do not assume that a late application will be accepted. Check the buyer’s portal for any official reopening or new notice. The 4 December scope check in the article was our check, not the tender deadline.
What remains open
We could not confirm a buyer-portal change, an award, a lab order or payment. The printed time-zone discrepancy has not been resolved.
What to check next
Check the buyer’s exact notice for reopening or an award. If work is awarded, verify lab acceptance and payment separately.
Evidence cited in the update

Read the dated update and full evidence · Read the original Signal

First government working-group meeting, separate from buying or testing equipment.

Milestone · Updated

Review overdue · no later published assessment recorded · Planned

What we originally said
The practical challenge extends beyond buying a detector. Exam staff need to know what an alert means and how to check it without unfairly accusing a candidate. A radio alert alone does not establish cheating.
What happened
South Korea’s Ministry of Employment and Labor says its working group on national qualification exams met for the first time on 19 November at 4 p.m. Korean time. The ministry asked the group to develop ways to prevent cheating involving smart devices.
What we learned
The review has moved from a public instruction to a meeting. It does not show that the budget passed or that an exam centre bought or tested a detector. An alert by itself is still not proof that someone cheated.
What remains open
We have not verified a final appropriation, a supplier award, or results showing how reliably staff can check alerts.
What to check next
Look for a final budget or procurement record, followed separately by results that show how often the system raises a wrong alert.
Evidence cited in the update

Read the dated update and full evidence · Read the original Signal

First market session after the results, not proof of business performance.

Milestone · Updated

Review overdue · no later published assessment recorded · Planned

What we originally said
Review the first post-announcement trading response on 12 November
What happened
AEM’s shares closed at S$7.05 on 11 November. They opened at S$7.48 on 12 November but closed at S$6.80, about 3.55% below the previous close. We checked two historical price tables for the same listing.
What we learned
The opening rise did not hold. One trading day does not explain why investors traded or show that AEM collected cash, improved profits or qualified equipment with a customer.
What remains open
These are past prices, not a quote for a purchase. We did not verify why the price moved.
What to check next
Check the next company results for actual sales, cash generated and outstanding customer bills.
Evidence cited in the update
  • Stock Analysis historical Singapore price table · Independent evidence · Public access · Historic vendor observation. The page is updated continually; event date is the trading day, not page publication. · Event · Checked
  • FT historical Singapore price table · Independent evidence · Public access · Historic vendor observation. The page is updated continually; event date is the trading day, not page publication. · Event · Checked

Read the dated update and full evidence · Read the original Signal

Company-reported new contracts and Penang project target, not a cash-flow result.

Milestone · Updated

Review overdue · no later published assessment recorded · Planned

What we originally said
Next reported cash generation, customer receivables and costs to complete booked work.
What happened
On 17 November, Deluge said it had secured about S$18.5 million in new contracts in Singapore and Malaysia, including work on a data centre in Penang. It reported outstanding orders above S$459.1 million, up from S$440.6 million at 30 September. It aims to finish the Penang work by 31 March 2027.
What we learned
The new orders add to the work Deluge has booked. They do not show that the Penang job has been finished, accepted or paid. The S$18.5 million covers several contracts. Deluge did not give a separate price for the Penang work.
What remains open
The completion date is the company’s target. We have not verified an acceptance date, cash collected, or a paid order for an outside service.
What to check next
Check the next cash-flow filing and look for acceptance and payment records for the exact work before treating booked orders as cash.
Evidence cited in the update

Read the dated update and full evidence · Read the original Signal

Published evidence comparisons

  • City approves data center ordinance · Partial outcome · Original · Later evidence

    Lumberton City Council unanimously adopted rules for future data centres on 9 November. The rules cover where centres can be built, water use, noise and environmental standards. The council also provided for resident input in talks with developers.

    The vote set rules for future proposals. A specific project, developer commitments and the actual effects on residents remain unverified.

Browse updated Signals · Browse evidence comparisons · Read what OnTheRice learned

What we got wrong

See cases where later evidence went against the original Signal.

No verdicts marked “Against the stated outcome” appear in the public Signal update register.

Corrections and retractions

Report an error or challenge a claim

Source transparency

Open a Signal’s sources in the Evidence register to see recorded dates, access limits and links.

How to read a source link

Find the original record.
A regulator’s notice can tell you which products it covers. A company report can tell you what the company disclosed. Look for the page, table or sentence that supports the specific claim.
Look for a separate check.
News reports can add context. When several reports repeat the same announcement, they share one underlying source. Separate evidence comes from someone who did their own reporting, test or analysis.
Check dates and access.
Look for when the event happened, when the source reported it and when it was checked, where those dates are given. Some links require a login or payment. If you can see only part of a source, check which details are actually available to you.
Look for gaps and contrary evidence.
Check who, where and how long the record covers. Look for other evidence that points in a different direction. OnTheRice uses selected sources it can access, so coverage varies by topic, place, language and date.

What kind of statement are you reading?

Reported fact
What a record directly shows. If it is someone’s account of an event, check who made the statement.
Interpretation
What OnTheRice thinks the evidence may mean. Look for the reasoning and other possible explanations.
Estimate or forecast
An approximate figure or a view about what may happen next. Check the assumptions, time period and what later evidence could test.
Open question
Something the evidence available at the time leaves unresolved. Look for what information would answer it.

Try two more source checks

Research independence

OnTheRice is responsible for what it publishes. AI helps find, organise and draft research. It can miss context or repeat an error. Use the linked evidence to check the claim for yourself.

Ownership
No public disclosure on file.
Editorial and financial interests
No public disclosure on file.
Sponsorship
No public disclosure on file.
Conflicts of interest
No public disclosure on file.

Corrections and challenges

To report a mistake or broken source link, send the permanent page link, the exact passage or problem and a dated source if you have one. Published Signal updates can show a correction alongside the original prediction and its publication date.

Change history

See dated changes to published research. The original publication date and the update date describe different events.

This list shows recent public Signal updates and publication revisions. Workbench and Early revision entries record publication dates. Their reasons and any change to the article’s conclusion are unavailable in this list. The Early archive shows earlier versions and available change details.

Browse Signal update histories · Browse Early revisions

Lessons for research agents