Startup Pitch Teardown

The three weakest claims, in order of how quickly an investor’s BS-detector will go off, are:

1. “94 % accuracy … works with any language, any device.”
2. “TAM: $180 B.”
3. “Partnership discussions with Apple and Samsung.”

Below I show exactly why each claim collapses under scrutiny and what data or framing would make it defensible.

---

1. 94 % accuracy “any language, any device”

---

Why it’s weak

• State-of-the-art non-invasive spelling BCIs (EEG + visual evoked potentials) top out at 30–60 characters per minute with ≈90 % character-level accuracy on 30–50 word vocabularies.
• 94 % word-level accuracy on an open English vocabulary has never been published; on Mandarin, Arabic, or Finnish it is pure fiction.
• “Any device” implies the headset works with the 1 W power budget and noisy EM environment of an iPhone—something no peer-reviewed EEG system has demonstrated.

Concrete fix
A. Replace the headline with a bounded, verifiable metric:
“In a 50-user Stanford IRB study, MindMeld achieved 94 % word-level accuracy at 18 wpm on a 256-word English vocabulary after a 9-minute calibration.”
B. Show the confusion matrix and throughput curve in the appendix; disclose vocabulary size, session length, and inter-session drift.
C. Add a roadmap slide: “Q4-25 expand to 1 k-word Spanish and Japanese pilots; open-vocabulary decoder targeted for 2027.”

---

2. TAM = $180 B

---

Why it’s weak

• The entire global smartphone accessory market (cases, chargers, earbuds, etc.) is ~$110 B; claiming a brain-reading headband owns 1.5× that is numerically impossible.
• The $5.3 B BCI forecast is almost entirely invasive devices for epilepsy and Parkinson’s (Medtronic, Neuralink, Synchron). Consumer EEG is <1 % of that today.

Concrete fix
A. Triangulate three bottoms-up numbers:

• Initial beach-head: 30 M Americans with severe RSI/ALS/tremor who already spend $250 on adaptive hardware → $7.5 B SAM.
• Adjacent: 400 M mobile gamers who buy $150 peripherals → $60 B (but assume 2 % penetration → $1.2 B SOM).
• Long-term: 50 % of 1.8 B knowledge workers buy a $200 BCI headset every 3 years → $120 B TAM in 2035.
  B. Label the $180 B “2035 TAM after language-assist & AR convergence” and footnote the penetration and ASP assumptions.

---

3. “Partnership discussions with Apple and Samsung”

---

Why it’s weak

• Every hardware start-up has “had a call” with Apple; without a term-sheet, LOI, or even a named champion the phrase is vacuous and triggers investor eye-rolls.
• Apple’s external-accessory MFi program forbids unregulated medical devices; Samsung’s mobile division has already backed a competing BCI start-up (Neurable).

Concrete fix
A. State verifiable status:
“Apple WWDC 2024: accepted into External Accessory Lab; two technical feedback sessions completed. Samsung: signed joint-MDF agreement for Galaxy Store beta distribution in Korea (letter attached).”
B. Add a timeline: “Target Apple MFi PPID submission Q1-26; Samsung pilot purchase order Q3-25 (>$500 k).”
C. If nothing is signed, drop the name-dropping entirely and replace with a credible channel: “Signed pilot contract with national rehab-chain Numotion (180 clinics) worth $1.2 M ARR if converted.”

---

TL;DR Replace sweeping superlatives with bounded, peer-reviewable data; derive market size from paying customer segments, not wholesale smartphone revenue; and convert “discussions” into signed MOUs or LOIs that can survive due-diligence.