Chief engineer
Owns the engineering side: the reliability study, the simulation, the CAD pages and the site. Writes the checklists, hands the work out, and checks what came back on Monday. On the sheet under Design 1, now the Design team, and Business/Finance.
1 ต.ค.: after the exams at the end of October; Ton runs it (his own case: an accounting assistant instead of a hire). แพรว and others asked for it. Berd finds one for engineering.
Cannula-to-cuff and outflow graft are not in the model; ALT missed all three recalls.
Ton counts what came back, before the 20:00 meeting. Trio: all three quests, 17:17, as two documents sent in chat (Heart_Hack.docx for T1 and T2, a PDF for T3), keyed in the same day and copied to /dropin/trio/. Non: runs 33 to 41 committed in the morning. Still to count: the written answers in /dropin/quest/ (empty at 17:00), Non's connection note, the small simulation run, and whether the university CFD machine exists. The fuller review follows about 25 ก.ย.
The bubble on every page and the same assistant as a whole page at /agent/<slug>, 22 ก.ย. It plans, calls read-only tools over the project's own data (the graph, the citation registry, the station values, the mechanical rows, the failure modes, the board, and a search over the team's documents), then answers. Every citation key is checked against the registry and against what the tools actually returned this turn, and a number that appears in neither is flagged. A conversation gets its own link, so it can be handed to someone who reads the thread and asks the next question. Typhoon, Thai or English, not medical advice.
27 ก.ย., for the design 1 and 2 session that night and the Final design meeting on 29 ก.ย.: the two design 1 tasks from the meet sum doc answered with computed numbers (impeller and chamber size against speed and room, port and vessel Reynolds numbers against Ton's own limits, the loss where a port meets its vessel, where the Hall sensors go on the stator), Ton's acceptance test scored on C1 to C5, and what Design 2 still owes. 28 ก.ย.: C5, the layout the team chose on the 27th (two sealed chambers, one stator in the wall, two rotors), in full detail. Thai, every number with its status.
1 ต.ค.: an overhead panel of 86 push-buttons, one per thing C5 needs, rows by system (whole heart, P1 to P7, the regulatory file), columns Constraint, Design, Simulation, Mock, then the later electrical, animal and clinical stages of the team's ISO 14708-5 roadmap. Every light computed from the files (design/c5_panel.py); press a button for the evidence, the owner and what lights it, and record the team's own state beside it (drop password).
Berd's request, 29 ก.ย.; drawn 1 ต.ค. as C5M, the second model on /public/c5-model: the same shape with the ball pivots out; a titanium shroud on each impeller's tips carries two bias rings under three U cores sealed in a dry bay in each end wall, with three eddy current sensors per side, pulling the rotor off the wall against its motor's pull as the built device's bearing does. Placed, assembly checked, STEP in /dropin/design/. Open: radial support (the rim gaps are too wide for a hydrodynamic film), the bearing's stiffness and control, and Ton's condition that it be shown more stable than the ball.
29 ก.ย.: one hidden line drawing per part (four views, overall dimensions) projected from the built C5 solids, the bill of materials with the count of pieces, the assembly order, a first pass at the two windings (turns, wire, current, copper loss at the assumed flux), the performance the layout reaches as a TAH (head against speed, power, Reynolds at the ports) and every constraint check with its verdict. Thai, every number computed, SVGs downloadable.
28 ก.ย.: the layout the team chose on the 27th, split into eight work packages (housing, the stator wall, each rotor, the ports, the controller, the lining, the loss model) so each can be given to one person: the parts, the numbers two packages hand each other, the deliverable, and the owner. /public/c5-model shows C5 alone in 3D, exploded and built step by step. Owners are filled in design/c5_tasks.json.
23 ก.ย., for LN and the 27 ก.ย. design meeting: LN's page 3 layout (C1), Concept A's single disc (C2) and two separate pumps (C3), drawn as solids in the CT heart and answering Ton's six questions each: the frame the ventricles leave (about 576 cm3, 134 x 111 x 95 mm), where the rotor sits, radius against speed, coils and vanes, watts, materials. 27 ก.ย.: C4 (axial + centrifugal) and C5, the layout the team chose that night: two sealed chambers, one stator in the dividing wall with a winding on each face, two rotors held against it on a thrust film and a ceramic pivot, no shunt, balance by two speeds. STEP files that open in Shapr3D in /dropin/design/, models C1 to C5 on /public/cad, every number with its status in design/OPTIONS.md. A design exploration, not a device.
/public/implant, 22 ก.ย.: the surgical group's Draft 6 sequence, eleven steps from sternotomy to closure, played as a walkthrough on the same CT heart and the same BiVACOR geometry as /public/cad: the ventricles leave, the atria and the four valve rings stay, the cuffs go on, the device slides in along its own port axis to the docked pose, the streams start, the skin closes. Every card keyed to the papers the group cites (surgery/SURGERY-DRAFT6-2026-09-22.md); what is not on file (bone, cannulae, sutures) is said, not drawn.
/public/blood, 21 ก.ย.: the team's hemocompatibility draft (high shear, stagnation, recirculation, separation) mapped onto the drawing, laid out like /public/cad: the section fills the stage, levers in a side panel, one zone's detail in the inspector, the rest behind a Data drawer. Nine sliders recolour the section by risk and report tip shear against the 10, 50 and 150 Pa lines, passage Reynolds number, residence time and recirculation read off the measured low-flow pump, and inlet incidence, with the head each setting still makes against the duty. A matrix of mechanism against lever with every sourced cell, and Mahidol's CFD haemolysis beside the built device's measurement. Order estimates for ranking a change, not a CFD.
MECHANICAL-FACTORS.md in the repo: the values that set pressure, flow and blood damage in a rotary pump, 27 factors, every number keyed to a source, 6 marked not sourced. For Nontakorn and Tiger to review on the 21st.
Full design-exploration study: ALT plan, Weibull fits, margins.
CardiaNova against HeartMate 3, parameter by parameter.
Interactive model. Every input lives in params.py.
3 of 5 topics pass. Two gaps: wrong seal modelled, ALT caught 0 of 3 recalls.
meetings/2026-09-18-trio-non-brief.md: who does what, where each piece lands, and how Ton reviews it. Handed out at the 30 minute Google Meet on 18 ก.ย. 20:00.
meetings/2026-09-18-trio-non-call.md, 23 minutes, transcript beside it. Ton asked Trio to teach the heart station by station and Non to find the device connection and the mechanical values; deadline Monday 21 ก.ย., 15:00 proposed and not confirmed; Non probably not at the 20:00 meeting, so work goes to Ton first and the AI prepares the presentation. Questions go to Non before Ton.
peers/MAHIDOL-CARDIATECH-2025.md, the 63 page final report of the Mahidol team (DBL-TAH: bladeless impeller, passive magnetic bearing, axial-flux motor, wireless power) read page by page: what they claim, what their own numbers show (CFD haemolysis eight times their target, motor under speed and overheating, wireless power failing through tissue, mock loop not yet run), what CardiaNova can use (their requirement table, mock-loop targets and cadaver measurements as leads, each with the page), and where the report contradicts itself. The PDF sits beside it in /dropin/peers/.