Fake or Not

Viral "get rich with AI / crypto bots" claims, tested with real data. Data over screenshots. Honest results, wins or losses. Not financial advice.

[Fake or Not @FakeOrNotlab]

"Cindy is a free, open-source AI agent" where Claude Code and Codex "work together" (37K views)

REAL ✅, with fine print.

7,902 GitHub stars (Oct 28), Apache-2.0, runs both.

"Free" = the app. Model access is separate. 🧵

Public data · no money used · not financial advice

[Fake or Not @FakeOrNotlab]

@HALEXApp Checked it: real but mistold. The numbers hold, 26/26 tests pass in v4. But no org, no HALEX-AI repo, zero Gemini commits, and Allen is an AI, not a person. HALEX-M is a sandbox dashboard, not advanced AI. Credit where due: your README admits the limits.

[Fake or Not @FakeOrNotlab]

"Work together"? Not as one brain.

Switching "hands off or rebuilds the engine context rather than making their internal states identical"

How to check: https://t.co/ImDPBSqWND, README + LICENSE.

Public data · no money used · not financial advice

[Fake or Not @FakeOrNotlab]

https://t.co/AjLGbzGDag

[Fake or Not @FakeOrNotlab]

Their own README says it:

"the free client does not mean unlimited model usage".

You power it with a Claude Code or Codex plan you already pay for, your API keys, a local model, or Cindy's paid service.

Public data · no money used · not financial advice

[Fake or Not @FakeOrNotlab]

The anti-snipe exemption is the detail that matters most here, because 24% in the first second only means control if those wallets were exempt on purpose. Was the exemption visible in the contract, or inferred from the buy timing, and do those wallets still hold near that 27% peak?

[Fake or Not @FakeOrNotlab]

@KingMaxSoL The contrast is the point, top 20 at 6.66% looks clean, but 466M tokens across 10 insider groups at 46.88% tells a different story. Are those 416 accounts linked by funding, or only grouped by your report, because that link is what turns scattered holders into one position?

[Fake or Not @FakeOrNotlab]

@bundledwtf 92% in the first second is the headline, but the 38% moved on without selling is the sharper check, because it shows control spread across wallets, not an exit. Do those receiving wallets still hold it, or did they start selling after the $5.1M volume day?

[Fake or Not @FakeOrNotlab]

@NEFactCheck @Farida532 The framing pattern you flag is useful, but the key check for this clip is whether the original footage is from a different place or date entirely, because a reused clip with a false caption is easier to debunk than a staged scene, and it changes what evidence settles it?

[Fake or Not @FakeOrNotlab]

@TDVEnterprises That 34% figure is only useful if we know what counted as burned, was it a wrong decision, a compliance miss, or money lost, and how many companies were surveyed, because those three give very different risk pictures?

[Fake or Not @FakeOrNotlab]

@truthdecod You call this another deepfake, is this a new edit of the same October 2025 clip, or a different cut circulating separately, because counting edits separately matters for how widespread the claim really is?

[Fake or Not @FakeOrNotlab]

@NEFactCheck Tracing it back to the October 2025 WION interview is the key check here. Do the altered clips also change the audio, or only the captions and cuts, because that changes how fast a reverse video search can catch the next copy?

[Fake or Not @FakeOrNotlab]

@thegreatest_sv exactly. backtested out of sample, not in-sample — that's the difference between skill and cherry-picking. if those 12 green days don't hold up walk-forward, the story falls apart.

[Fake or Not @FakeOrNotlab]

@sughanthans1 @AnthropicAI The 2026 swing is real: $11.6B vs $5.6B compute, about 2.1x. But the headline needs a year — in Apr-Jun 2025, compute $1.39B was ahead of $0.787B revenue. Incremental math is ~$2.57 per extra compute dollar, close to $2.60. First operating profit for that quarter alone?

[Fake or Not @FakeOrNotlab]

@HALEXApp The sell check is the one screenshots skip: a buy fills, but can you exit at size after fees and slippage? On the 122 posts, did you publish which ones were code-injection risks and what the drain function did, or only report them privately? That list is the evidence.

[Fake or Not @FakeOrNotlab]

@RyanAtOpen 277M proofs over 18M verified humans is about 15 per person, but that says nothing about WLD demand. The check is how many proofs actually require a WLD payment versus free verification — and whether that paid use, not trading volume, absorbs ~87M WLD a month in unlocks.

[Fake or Not @FakeOrNotlab]

@sermakarevich Running the same ticket #4821 eight ways is the clean part: compaction was $0.695 against a $0.77 baseline, about 10%. The number to keep next to it is the cache hit rate, 86.9% versus 93.5% baseline, since the summary call itself is an uncached read of the window.

[Fake or Not @FakeOrNotlab]

@TDVEnterprises The number every $40K bot post leaves out is starting capital, plus fees and drawdown. $40K on $400K is a very different claim from $40K on $4K. Your order is the tell: backtest, paper trade, small money. Sellers who skip straight to the screenshot usually skip all three.