AskNoema

Verify what you watch, as you watch it

The platform where you research and verify your own feed. Multi-agent AI checks the claims inside social media and trending news in real time — without ever pulling you out of the stream.

Get pre-release access See how it works ›

No product access is public yet.

EPISTEME RUNTIMEReasoning trace01CLAIM EXTRACTED“They finished the Great Pyramid in about 20years.”02PROVENANCE TRACEDFigure originates with Herodotus, c. 450 BC —roughly two millennia after the build.03INDEPENDENT CORROBORATION✓Turin Canon — Khufu credited 23 years✓Wadi al-Jarf papyri — casing stone in transit, regnalyear 26✓Giza settlement — provisioned, non-servileworkforce04SOURCE SPLIT, NOT ACCEPTED WHOLE✓Duration — corroborated✗50-year reign — contradicted✗Enslaved labour — contradicted05FEASIBILITY CHECKED2.3M blocks ÷ 23 yr≈ 285 / day≈ 1 per 2 min daylight06RESIDUAL UNCERTAINTY LOGGEDReign length itself is disputed — 23, 27, or longer.AskNoemaDid the Great Pyramid really take about twenty years to build?Confirmed · 0.8912 sources · 4 primaryBroadly yes — though the number reaches us from an unreliable narrator whohappened to be right.The twenty-year figure comes from Herodotus, who collected it from Egyptian informantsaround 450 BC, some two thousand years after the pyramid was finished @histories.ii. For mostof the modern era that was the only testimony we had, and a single late source is thin ground.Two independent lines have since converged on it. The Turin Canon assigns Khufu a reign oftwenty-three years @turin-canon, bracketing the window. And in 2013 Pierre Tallet’s excavationat Wadi al-Jarf recovered the logbook of an inspector named Merer — the oldest inscribedpapyrus known — recording his crew ferrying Tura limestone to Akhet-Khufu in the twenty-sixthregnal year @tallet-2017.That cargo matters. Tura limestone was casing stone, the final cladding. Work was stillunderway at the very end of the reign — which rules out the shorter ten-year estimates someEgyptologists had proposed.The arithmetic is demanding but not impossible: roughly 2.3 million blocks across twenty-threeyears is about 285 placements a day @giza-project.org.OpenThe reign length is not settled. Manetho gives Khufu sixty-three years, Herodotus fifty, andStadelmann has argued the Canon’s 23 is a scribal inversion of 32. The construction window isfirmer than the reign that contains it.EVIDENCE12 resultsTmuseoegizio.it › turin-canonRoyal Canon of Turin — Fourth Dynasty columnRamesside king list attributing 23 years and some months to Khufu;compiled from earlier annals.PrimaryCited 340+Wamers.hypotheses.org › papyrus-jarfLes papyrus de la mer Rouge I — Le Journal de MererTallet, 2017. Daily logbook of limestone runs from Tura toAkhet-Khufu, dated to regnal year 26.PrimaryPeer-reviewedNnature.com › s40494-025-02018-wConstruction of the Great Pyramid with counter-weightedrampsNotes the ~20-year duration aligns with the Canon, and that Wadial-Jarf rules out ~10-year models.Peer-reviewed2025Aaeraweb.org › giza-plateau-mappingWorkers’ settlement: bakeries, breweries, tombsLehner. Provisioned labour town indicates a paid, organisedworkforce, not enslaved.PrimaryRnationalgeographic.com › red-sea-scrollsThe Red Sea Scrolls: sealed galleries at Wadi al-JarfHundreds of papyrus fragments recovered in 2013; the oldestinscribed papyri known.Reported EPISTEME RUNTIMEReasoning trace01CLAIM EXTRACTED“They finished the Great Pyramid in about 20years.”02PROVENANCE TRACEDFigure originates with Herodotus, c. 450 BC —roughly two millennia after the build.03INDEPENDENT CORROBORATION✓Turin Canon — Khufu credited 23 years✓Wadi al-Jarf papyri — casing stone in transit, regnalyear 26✓Giza settlement — provisioned, non-servileworkforce04SOURCE SPLIT, NOT ACCEPTED WHOLE✓Duration — corroborated✗50-year reign — contradicted✗Enslaved labour — contradicted05FEASIBILITY CHECKED2.3M blocks ÷ 23 yr≈ 285 / day≈ 1 per 2 min daylight06RESIDUAL UNCERTAINTY LOGGEDReign length itself is disputed — 23, 27, or longer.AskNoemaDid the Great Pyramid really take about twenty years to build?Confirmed · 0.8912 sources · 4 primaryBroadly yes — though the number reaches us from an unreliable narrator whohappened to be right.The twenty-year figure comes from Herodotus, who collected it from Egyptian informantsaround 450 BC, some two thousand years after the pyramid was finished @histories.ii. For mostof the modern era that was the only testimony we had, and a single late source is thin ground.Two independent lines have since converged on it. The Turin Canon assigns Khufu a reign oftwenty-three years @turin-canon, bracketing the window. And in 2013 Pierre Tallet’s excavationat Wadi al-Jarf recovered the logbook of an inspector named Merer — the oldest inscribedpapyrus known — recording his crew ferrying Tura limestone to Akhet-Khufu in the twenty-sixthregnal year @tallet-2017.That cargo matters. Tura limestone was casing stone, the final cladding. Work was stillunderway at the very end of the reign — which rules out the shorter ten-year estimates someEgyptologists had proposed.The arithmetic is demanding but not impossible: roughly 2.3 million blocks across twenty-threeyears is about 285 placements a day @giza-project.org.OpenThe reign length is not settled. Manetho gives Khufu sixty-three years, Herodotus fifty, andStadelmann has argued the Canon’s 23 is a scribal inversion of 32. The construction window isfirmer than the reign that contains it.EVIDENCE12 resultsTmuseoegizio.it › turin-canonRoyal Canon of Turin — Fourth Dynasty columnRamesside king list attributing 23 years and some months to Khufu;compiled from earlier annals.PrimaryCited 340+Wamers.hypotheses.org › papyrus-jarfLes papyrus de la mer Rouge I — Le Journal de MererTallet, 2017. Daily logbook of limestone runs from Tura toAkhet-Khufu, dated to regnal year 26.PrimaryPeer-reviewedNnature.com › s40494-025-02018-wConstruction of the Great Pyramid with counter-weightedrampsNotes the ~20-year duration aligns with the Canon, and that Wadial-Jarf rules out ~10-year models.Peer-reviewed2025Aaeraweb.org › giza-plateau-mappingWorkers’ settlement: bakeries, breweries, tombsLehner. Provisioned labour town indicates a paid, organisedworkforce, not enslaved.PrimaryRnationalgeographic.com › red-sea-scrollsThe Red Sea Scrolls: sealed galleries at Wadi al-JarfHundreds of papyrus fragments recovered in 2013; the oldestinscribed papyri known.Reported

The problem

The future of social networks is social intelligence

AskNoema is built for active feed research — powered by multi-agent AI that verifies media claims and trending news in real time.

Our solution

Verification that travels with the content

Instead of cross-checking tabs or falling for online misinformation, AskNoema lets you consume trending social content alongside verified, instant multi-agent AI claim breakdowns, source verification, and deep research.

@ancient.builds · 2hCasing stone for the Great Pyramidwas quarried at TuraConfirmed · 0.948 sources@deep.time · 3hThe Sphinx predates the FourthDynasty by millenniaContested · 0.4211 sources@museum.notes · 5hMerer's logbook is the oldest inscribedpapyrus knownConfirmed · 0.976 sources@fastfacts · 7hConstruction required a standing armyof slavesFalse · 0.0615 sourcesAnalysed71@ancient.builds · 2hThe Great Pyramid took about twenty yearsto buildConfirmed · 0.8912 sources@viral.history · 4hA workforce of 100,000 men labouredyear-roundMisleading · 0.319 sources@thread.reply · 6hModern cranes could not lift the heaviestgranite blocksFalse · 0.1114 sources@sourcecheck · 8hHerodotus visited Giza roughly two millenniaafter the buildConfirmed · 0.967 sources@openarchive · 9hRamp geometry remains an unsettledquestionContested · 0.5010 sourcesQueued@fieldnotes · 10hLimestone was floated on seasonal floodchannelsConfirmed · 0.889 sources@replyguy · 11hThe blocks were poured concrete, not cutstoneFalse · 0.0413 sources@archive.eg · 13hWorker rations included bread and beer dailyConfirmed · 0.926 sourcesThreads by @asknoemaCCharlotte3:32 PMSaw a clip claiming we couldn't rebuild the Great Pyramid today, even with cranes andlasers. Is that actually true?DDiana3:34 PMWe could. We just wouldn't. Nobody is funding twenty years of stonework for onebuilding — it's a question of will, not capability.EElizabeth3:37 PMThat's the part that unsettles me. We can look up how anything was made and buildalmost none of it ourselves. The knowing has drifted outside us.CCharlotte3:39 PM@asknoema is the twenty-year figure actually established, or is it Herodotus repeatedfor two thousand years?AskNoema3:39 PMTWN+9SourcesConfirmed · 0.89Herodotus is where it starts, but he is no longer the only evidence. The Turin Canoncredits Khufu with 23 years, and Merer's logbook — excavated at Wadi al-Jarf in 2013— records limestone still moving to Giza in his 26th regnal year. Two independentlines, one window.246140836 @ancient.builds · 2hCasing stone for the Great Pyramidwas quarried at TuraConfirmed · 0.948 sources@deep.time · 3hThe Sphinx predates the FourthDynasty by millenniaContested · 0.4211 sources@museum.notes · 5hMerer's logbook is the oldest inscribedpapyrus knownConfirmed · 0.976 sources@fastfacts · 7hConstruction required a standing armyof slavesFalse · 0.0615 sourcesAnalysed71@ancient.builds · 2hThe Great Pyramid took about twenty yearsto buildConfirmed · 0.8912 sources@viral.history · 4hA workforce of 100,000 men labouredyear-roundMisleading · 0.319 sources@thread.reply · 6hModern cranes could not lift the heaviestgranite blocksFalse · 0.1114 sources@sourcecheck · 8hHerodotus visited Giza roughly two millenniaafter the buildConfirmed · 0.967 sources@openarchive · 9hRamp geometry remains an unsettledquestionContested · 0.5010 sourcesQueued@fieldnotes · 10hLimestone was floated on seasonal floodchannelsConfirmed · 0.889 sources@replyguy · 11hThe blocks were poured concrete, not cutstoneFalse · 0.0413 sources@archive.eg · 13hWorker rations included bread and beer dailyConfirmed · 0.926 sourcesThreads by @asknoemaCCharlotte3:32 PMSaw a clip claiming we couldn't rebuild the Great Pyramid today, even with cranes andlasers. Is that actually true?DDiana3:34 PMWe could. We just wouldn't. Nobody is funding twenty years of stonework for onebuilding — it's a question of will, not capability.EElizabeth3:37 PMThat's the part that unsettles me. We can look up how anything was made and buildalmost none of it ourselves. The knowing has drifted outside us.CCharlotte3:39 PM@asknoema is the twenty-year figure actually established, or is it Herodotus repeatedfor two thousand years?AskNoema3:39 PMTWN+9SourcesConfirmed · 0.89Herodotus is where it starts, but he is no longer the only evidence. The Turin Canoncredits Khufu with 23 years, and Merer's logbook — excavated at Wadi al-Jarf in 2013— records limestone still moving to Giza in his 26th regnal year. Two independentlines, one window.246140836

Capabilities

Three things the feed has never done

Context-locked overlay

Context-locked AI overlay and real-time social intelligence

Never get misled by a viral clip again. AskNoema analyzes content in real time, delivering instant source ratings, verified facts, and context-locked analysis without ever leaving your stream.

Research dossiers

Topic deep dives and interactive research dossiers

Instant academic depth for every video topic. Convert brief video mentions into cited, structured research dossiers. Explore primary sources, counter-arguments, and deep context without ever leaving your active session.

Source lineage

Source lineage and fact-checking

Trace every online claim back to its primary origin. Eliminate noise and unsourced opinion. AskNoema maps creator claims against real-time web indexes, peer-reviewed literature, and primary data streams — giving you full visibility into where facts come from.

A verified AskNoema response about the attack on Pearl Harbor, its three checked claims and cited sources, beside the field of pipeline stages that produced it.

The workflow

multi-agent workflow system between media and AI analysis output

Each subagent deploys its output to the next, emulating the multi-step verification system on verified media houses.

01 / 09

L0 media.intake L2 asr.whisper L3 vision.gemini L3 ocr.tesseract L3 record L4 claim.sonnet L5 verify.sonar L6 synthesis L7 display

media.intake L0 · Boundary

Media intake

The upload media is demuxed into an audio track, a sampled frame set and located text regions.

In
upload
Out
demuxed streams

asr.transcription L2 · Extraction

Transcription

Transcription agents runs to retrieve timestamped verbose data from uploaded media.

In
16 kHz mono PCM
Out
segments + word timings

vision.recognition L3 · Extraction

Vision annotation

Vision agents describes what each frame shows — subject, setting, figures on screen, documents held up to camera.

In
media frame set
Out
per-frame annotations

ocr.analysis L3 · Extraction

OCR · analysis

OCR systems reads the located regions verbatim, and catches exact wording that a vision models paraphrase.

In
text regions
Out
literal strings

Data Assortment L3 · Contract

Episteme Proprietary Agent

Transcript, annotations and OCR merge into a single contract-typed artifact, and every later stage reads and writes here rather than touching raw media. It is also where resolved claims are cached by fingerprint.

In
extractor outputs
Out
one typed artifact

claim.extraction L4 · Reasoning

Claim extraction

Agent separates propositions from everything else. Rhetoric, framing, opinion and jokes are set aside so the verifier only ever sees statements that can actually be true or false. Most of a clip is not a claim.

In
record
Out
checkable claims

web.verification L5 · Evidence

Web verification

Each claim goes to the open web with its own query set. Sources are weighed rather than counted — one primary document can outrank ten aggregators repeating each other, and a single source can be split.

In
claim set
Out
weighed sources

synthesis L6 · Assembly

Synthesis

Evidence is reconciled into a verdict and a veracity score. Disagreement is preserved rather than averaged away: where credible sources genuinely conflict, the output says contested and shows the split.

In
record + evidence
Out
verdict + veracity

display L7 · Assembly

Display

The response is content-locked to the video and timestamp it was asked about, then rendered. Nothing here is computed on the client — verdict, score and sources all arrive from the runtime.

In
response object
Out
feed surface

Be early to social intelligence

AskNoema is in active development at Ajir Research Labs. Pre-release access goes out in small groups, to people who actually research their feed.

Get pre-release access Read the research ›