I am blind or visually impaired: the story in one minute
Third Eye Kit

Research, Archive, Design review

This review answers the brief of the founder, Said Mohaddes Sadeqi, published as submitted: A research-grade platform site, the brief.

Board review: a research-grade platform site

The founder, Said Mohaddes Sadeqi, asked for the site to present the Third Eye Kit as a research-grade, open-source haptic platform, for an audience of admissions committees, researchers, and accessibility developers, in five sections: a new hero with a metrics banner, a psychophysics framework, an architecture and API section, an evaluation and co-design framework, and an open-ecosystem section. The brief is published beside this review: A research-grade platform site: the founder's brief.

The same four reviewers read it: haptic psychophysics, embedded systems and latency, human-computer interaction and accessibility, and program. The same rules apply as in the two earlier reviews: every claim against a source, each request adopted, amended, or declined, arithmetic shown where a number changes. Two things constrain this review from the start. The founder's own committee-readiness brief, sent twice, forbade a redesign and protected the site's honest register; and the board decided yesterday what the display channel is, where it sits, and how fast it can be. This review does not reopen those decisions. It applies them.

The verdict in one paragraph

Adopt the substance, decline the packaging. The site was missing exactly what sections two, four, and five ask for: one page that presents the program as research, with its psychophysical basis, its evaluation plan, its open-hardware commitments, and a first draft of the wire format developers will need. That page now exists at /research, built from the same data files as the rest of the site, and the navigation points to it. The kit's own words stay: feel the world before it touches you, modeled, nobody has proven it yet.

Scorecard

Section Verdict In one line
Context Amended The display band is beside the kit, on the back at 4 by 3, at about 60 and 130 ms, as decided.
1. Hero and vision Substance adopted, hero and banner declined as written Sensory substitution is the right frame; it gets a research page, not a new front door; the metrics are the true ones.
2. Psychophysics and ergonomics Adopted Already in the placement study; summarized on one page with the ego-centric rationale and its sources.
3. Architecture and pipeline Adopted, amended Two chains, one boundary; T.E.K.-Link 0.1 as a 20-byte packet carrying the 16-byte array.
4. Co-design and evaluation Adopted, one word amended The framework the site lacked; "clinical" becomes evaluation with ethics review.
5. Open ecosystem Adopted, amended not published

Of 24 checkable items, 19 enter TEK, 12 as written and 7 amended.

The context, checked against the record

The brief describes the kit as a 4 by 4 or 5 by 5 grid on the back or torso, driven by a phone over Bluetooth at under 35 ms, trained in games. Three of those four sentences were reviewed yesterday and stand only as amended: the grid is 4 columns by 3 rows on the upper back because the back tells vertical neighbors apart at about 66 mm and horizontal ones at about 31 mm (back acuity study 2024); the display is a reading channel beside the belt, collar, and ankle bands and never carries a warning; and the chain runs at about 60 ms from a game and about 130 ms from a camera, because the Bluetooth connection interval floor is 7.5 ms, 15 on iOS, and the camera alone is about 80 ms (Bluetooth Core 6.2 overview, Punch Through, inovex). The fourth sentence, games as a training ground, stands, with Connors et al. 2014 as its evidence. Everything below inherits these.

1. Hero and vision

The frame is right. Sensory substitution is the correct name for what the kit does: information from an artificial receptor reaches the brain through a human-machine interface, and the brain learns to use it in place of the missing sense (Bach-y-Rita and Kercel 2003). The kit is non-invasive, worn over clothing, and its founding experiment is Bach-y-Rita's 1969 back display (Nature). Adopted, as the opening of the research page.

The hero is declined. The founder's committee brief said, in his words, do not redesign, rebrand, or restructure. The hero's line, "Feel the world before it touches you", is the site's thesis, and it says in seven words what the proposed value proposition would say in forty. The research framing belongs on a research page one click away, which now exists, and in the navigation, which now names it.

The metrics banner is adopted with true numbers. The brief lists four: target latency under 35 ms, open-source hardware, spatial grid resolution, and battery and thermal efficiency. The first is declined for the third time; a target dressed as a metric on the front of a research site is the one thing a committee would catch. The other three are real and now appear on the research page from the site's own data: about 70 and 120 ms from the world to the skin for the warning path, modeled; 20 motors in the first build and 23 in the recommended map, from the zone catalog; 1 to 2 watts for the whole sensing stack against 10 to 20 for a camera rig, from the sensing study; the version 1 firmware under MIT and the version 2 dossier public. Every one carries its status word.

The register is declined. "Elite", "world-class", and "visionary" are what advertising says about itself. The site's register is plain and specific, and the founder protected it twice. The program's ambition is stated by its milestones, its reviews with recorded dissent, and its habit of writing down what it has not proven. The board judges that a reader from Harvard or MIT will trust that register more than the other.

2. Psychophysics and ergonomics

Adopted. The deep dive the brief asks for is the placement study, twelve sections with a source on every number (Where the body should feel). The research page now carries a one-screen summary of the findings the design rests on: 30 mm on the trunk at 91 percent (Johannesson et al. 2017); the back's 31 versus 66 mm; one head motor counted correctly 94 percent of the time against five at 5 percent (Dobrzynski et al. 2012); 8 waist motors at 98 percent against 12 at 74 (Cholewiak 2004); which skin keeps reading while walking (Karuei et al. 2011); three motors that cannot show a direction where four can. The fatigue strategy stands as decided yesterday: linear resonant actuators, bursts of 50 to 120 ms rather than steady vibration, and an advisory two-hour limit until the telemetry sets a real one, resting on the reversibility of vibrotactile adaptation (afferent adaptation).

Ego-centric against exo-centric, adopted with its evidence. The kit's first law, where it buzzes is where it would hit you, is an ego-centric mapping, and the torso is where that mapping is most natural: a vibration on the torso is read as a direction in the world, left is left and front is front, with a small consistent bias toward the midline of up to 10 degrees (van Erp 2005). An exo-centric picture, north up, would ask the wearer to rotate a map in the head at every turn; the board keeps it out of the warning bands and allows it only on the display band, when the wearer stops to read a room. The rationale is now on the research page in those terms.

3. Architecture and pipeline

Adopted, amended. The brief's pipeline, capture to edge filtering to binary packet to actuation, describes the display channel. The warning path is different and simpler, and the research page says both plainly: distance sensors to a reflex controller that keeps the nearest reading per column and fires a drop-off signature, to drivers and motors, with no radio in the loop; and the phone's path for reading a space, with its two honest timings. The isolation boundary between them is drawn, as the first review required. The animated version already lives on the engineering page (one hazard, world to skin).

T.E.K.-Link 0.1, adopted as a draft. The brief asks for a data schema for 16-byte haptic intensity arrays. The board decided yesterday that the SDK is the frame contract already published here with a wire encoding and a version. The research page now carries that draft: a 20-byte packet, four header bytes and sixteen amplitude bytes, which fits the default Bluetooth attribute payload so no fragmentation is needed. The header carries the version, the band and its flags, a sequence number, and a rhythm period; the sixteen amplitudes carry either the display band's cells or the warning bands' regions in a fixed order with spare slots. It is marked draft, licensed MIT, and open for comment, and its first sender is the By feel game on this site. It is not a standard until a receiver has run it at the bench.

4. Participatory co-design and evaluation

Adopted, and it was the largest gap. The site had milestones and a safety file but no stated way of knowing whether the kit works. The research page now carries the framework: localization accuracy, sensory adaptation over a session, detection through clothing while walking, reaction time from onset to step, obstacle contacts and completion time on a fixed course, false alarms per hour, sessions to criterion, perceived workload and comfort, and transfer from the game to a real route, each with how it is measured and at which milestone. The obstacle-contact and completion-time measures follow the studies the site already cites (Unfolding Space Glove 2022, Connors et al. 2014); workload uses the NASA task load index (NASA TLX).

Participatory design, adopted as commitments. The founder lost his sight in the war, recovered sight in one eye, and is still blind in the other; he built the first kit from inside the experience; the first generation was tested with blind and visually impaired people in Herat. The program now states its rules in public: blind and low-vision co-researchers take part from protocol design onward, are paid, and are credited as authors on the write-ups; the questions they ask define the metrics; no study past milestone G1 runs without ethics review by a partner institution's board.

One word amended. "Clinical evaluation" implies a medical device and a claim the program does not make. The kit is a mobility aid under evaluation. The page says "user evaluation with ethics review". The standards path the dossier names, ISO 13482, ISO 14971, IEC 60601, IEC 62304, belongs to the powered exoskeleton and stays there.

5. Open ecosystem and reproducibility

Commitments with status, adopted. The research page lists what is public today: the version 1 firmware under MIT with its wiring and bill of materials, the version 2 dossier and safety file, eight drawings, the studies, the Council's record, and the simulator's contract and code on this site. It lists what arrives with the first build: KiCad schematics and board files, STL and pattern files for the bands and the pod, and the measured numbers that replace every "modeled". Nothing is claimed as available that is not.

Plugin guides, amended. Unity and Unreal hooks and a screen-translation guide are promised for the future in the brief; the board publishes guides for things that run. The order is the one decided yesterday: T.E.K.-Link 0.1, the game as the reference sender, Luanti as the open sandbox, then Unity and Unreal as reference integrations, then guides.

Markdown to paste, amended. The brief asks for markdown to paste into the repository. The site's pages are built from data files, so a number appears once and cannot drift: the latency model, the zone catalog, the parts list. The research page is code that reads those files, and the words on it are the program's. The board sees no reason to give up the property that makes the site trustworthy.

Decisions taken

  1. A research program page exists at /research, replacing the redirect to the papers list, and the navigation names it Research.
  2. The hero and the front page stand.
  3. The metrics on the research page are the site's true numbers, each with its status word; the 35 ms target does not appear.
  4. T.E.K.-Link 0.1 is published as a draft with the packet layout above, MIT, open for comment.
  5. The evaluation framework and the co-design commitments are published as the program's rules.
  6. The brief is published as written, minus its opening role instruction, which is not addressed to the board.

Sources