I would choose a stock market research platform by trying to recover one specific research object, not by counting its features. Can you identify what was forecast, how it was combined, which date governs it and where the original edition survives?
I inspected that public reading path on iPulse AI's NVIDIA pages on 15 October 2026. The current page exposed a frozen, synthesized price path. The history page exposed seven published batches and separate research-document links. A fresh page label did not make the underlying September forecast a new October prediction.
I am iPulse AI's founder, so this is a documented inspection of our own public pages, not an independent platform benchmark. It reveals what the inspected route allowed me to establish and where further evidence remained necessary.
The question I fixed before comparing features
My inspection question was: can I recover and explain one dated NVIDIA forecast without confusing it with a current quote, another aggregation method or an investment result?
That question has a concrete endpoint. A satisfactory record identifies the NASDAQ-listed NVIDIA asset, forecast anchor, horizon, return basis, aggregation and historical edition. It does not need to produce a buy recommendation.
The observation was a public-page inspection. I did not buy a subscription, compare paid accounts or measure how quickly a novice user completed the task. The table records fields actually found in the captured pages, rather than assigning a cosmetic platform score.
| Reading task | Public observation on 15 October | Boundary |
|---|---|---|
| Match the instrument | NVIDIA / NVDA.NASDAQ | A company name alone would not identify every possible listing |
| Find the path's reference | Frozen forecast from 18 September 2026; anchor 222.27 USD | This is an archived input, not today's quote |
| Identify its construction | Synthesized consensus using normalized opinion weights | This is not the archived equal-voice quantitative diagnostic |
| Read its units | Price returns excluding dividends; 20 quarterly steps | It does not measure investor total return |
| Recover the history | Seven published batches; separate AI Consensus and AI Opinions links | Archive availability does not establish forecast skill |
Source: the captured NVIDIA page and forecast history. These observations describe that edition of the public interface.
A successful retrieval can still answer the wrong question
The synthesized path displayed a one-year price return of 9.67% and a five-year price return of 57.44%. Those are stored scenario readings from its September anchor, not observed investment gains.
Our separately retained NVIDIA case reports equal-voice quantitative diagnostics with modeled dividends. Its one-year annualized total-return value is 8.84% and its five-year whole-horizon value is 59.50%. The construction and income basis differ.
Those numbers should not be collapsed into a claim that the platform changed its mind. A weighted price path and an equal-voice dividend-inclusive diagnostic are different objects. The inspection succeeds only when the reader can state which one they recovered.
The practical requirement is therefore not a large feature list. It is enough context to prevent an attractive number from being assigned to the wrong proposition. The forecast arithmetic guide explains the unit conversions; the Consensus Score methodology separates quantitative ranking from weighted synthesis.
A refresh label needs a source date beside it
The current NVIDIA page's captured text showed a page update dated 14 October and deep analysis published 21 September. The history page showed a 15 October performance refresh beside a latest-price reference dated 18 September.
That combination is a reason to inspect the underlying dates, not evidence that a new quote or research run occurred on the refresh date. The public labels alone do not establish why that source reference remained September-dated. I would keep the source date in the receipt rather than turn the newer label into a freshness claim.
This is an important test for any platform. Can it show the price observation, evidence cutoff, research generation and later calculation as separate dates? If not, ask for the missing boundary before comparing the output with another service.
A successful page load is not the same as successful research recovery. You may find the forecast path but still be unable to reconstruct the original retrieval log or every source supporting the generated narrative. Those gaps must remain explicit.
What the history established
The captured history page listed seven batches. For the latest batch it displayed 14 opinion receipts and separate links to the frozen historical AI Consensus and AI Opinions documents. Earlier entries also exposed unavailable document or receipt states instead of implying uniform historical coverage.
That is useful because the inspection can fail honestly. A missing historical document cannot be repaired by showing today's consensus. A receipt that was never captured cannot be recreated merely by giving an older page a new date.
I would keep the history link with the research note, along with the exact inspected edition. For a later comparison, reopen that edition first. Then identify what changed in the newer object: evidence, price basis, aggregation, formula or eligibility.
The Forecast Library account explains a further distinction. Preserving or externally anchoring a payload supports a narrow integrity claim. It does not validate the forecast, its source completeness or its investment merit.
Turn this inspection into a selection decision
Different readers need different recoverable objects. A filing researcher may need an exact reported line item and its PDF page. A quantitative-rating researcher may need factor definitions and rating history. An analyst-service reader may need the dated recommendation and thesis updates. A forecast researcher needs the original path and its later outcome rules.
Use the seven-tool comparison to identify those different output types. Then run a relevant retrieval task on the products you can actually access. Do not force a chart assistant to pass an unrelated archival-forecast test or treat an unavailable paid feature as evidence that the provider lacks it.
For each trial, preserve the question, available access tier, observed steps, output and remaining gap. Here the useful result was a readable September weighted path connected to an archive, with date and construction boundaries still requiring care. That is a more informative selection input than a count of AI features.
Evidence and limits
The observations and page captures are from 15 October 2026; source and internal links were reopened on 16 October before this revision. The table deliberately preserves the inspected edition rather than presenting it as a new live-price snapshot. This is one deliberate inspection of the owned NVIDIA route, not a representative comparison of platforms or users. The original 383-equity paired-horizon analysis remains in the preserved prior edition; it is not claimed as a new trial result. These archived scenarios do not establish current market prices, calibration or investment performance. The original publication date is retained separately from this material revision.




