AI research engine
Research that
shows its work.
Parallax gathers sources on any question, reads them, and writes a report where every claim is cited and every gap is named. So you can check the answer, not just trust it.
The standard
An answer is only as good as what stands behind it.
Closed book
Parallax writes only from the sources it retrieved for your question. It does not fill gaps from memory, however confident it might be.
Cited inline
Each claim carries its source right where it is made. Click a citation and you land on the source it came from.
Gaps named
Thin evidence, a lone source, an outdated figure, two sources that disagree — each one is flagged instead of smoothed over.
The report
Built to be checked.
Findings are rated by the strength of their evidence. Inferences are labelled as interpretation. What the sources couldn't answer is listed, not hidden.
Is sodium-ion storage ready for the grid?
Sodium-ion storage is operating at utility scale in pilot plantsS1, but cell costs have not yet fallen below the incumbent chemistryS2S4. Interpretation: technically proven, not yet economically settled.
- 01
A grid-connected plant has reported round-trip efficiency above 90%S1.
Strong - 02
Analysts dispute how soon costs reach parityS2S5.
WeakContested
No retrieved source reports long-term degradation data beyond pilot scale.
How it works
From question to cited report in three steps.
- 01
Ask
Pose a topic or a question. Add context — audience, geography, timeframe, the decision you're facing — to narrow the scope.
- 02
Gather
Parallax searches the web for sources. You can also paste in your own material, or switch the web off and use only that.
- 03
Read
You get a structured report with a bottom line, the findings behind it, and an honest account of what the evidence can't tell you.
A quick scan from a handful of sources.
A balanced report. The default.
The widest search and the fullest report.
Inside every report
More than a summary.
Bottom line
A short answer up front, with a stated confidence level and the reason for it.
Key findings
Each finding rated strong, moderate or weak by the evidence that supports it.
Contradictions
Where sources disagree, both positions side by side with the likeliest explanation.
Comparison
For either-or questions: options weighed against the same criteria, then decision guidance.
Gaps and open questions
What the sources didn't cover, and the searches worth running next.
Source credibility
Every source used, ranked into tiers from primary research down to blogs and forums.