Timeline¶
What happened, in order, with links to the evidence. The narrative is in the blog.
How the work is done¶
- Back up before any write. Nothing is written to any flash until a dump exists that was read twice with the same hash.
- The working second unit is read-only. It's the reference, not a donor.
- Mains. The power board is never probed while plugged in. Logic runs from a bench supply.
- Every claim carries a confidence (
confirmed/likely/guess) and a source. - No firmware is distributed, ever, in any form.
2026¶
| Date | What |
|---|---|
| 28 or 29 Oct | A firmware update started from the Anova app stalls at 75 % and stays there overnight. |
| 29 Oct | After a power cycle the unit is dead: no beep, no display, no LEDs. Support ticket opened. Advice: leave it plugged in overnight. |
| 30 Oct | Still dead. |
| 1 Nov | Anova: out of warranty, 10 % off a new unit. Four things asked in return: a firmware image with recovery instructions, a replacement main board at the owner's cost, a mail-in reflash, or a refurbished unit. Awaiting a reply. Incident, the exchange word for word. |
| 1 Nov | Head opened, main board photographed. Chips identified: U1 TI CC3220SM2A (Wi-Fi MCU, no internal flash), U15/U16 XMC XM25QH64A (8 MiB SPI flash, the dump target), U7 TI CC2640R2F (BLE MCU). Same board layout as the 2018 FCC sample; the debug parts aren't fitted. Main board. |
| 1 Nov | What the CC3220S security means: production mode disables JTAG, the file system is bound to the chip, the application image is signed. The working unit's flash can't be cloned; the chip's UART bootloader is still open. Bench setup. |
| 1 Nov | Recovery order fixed: look, map, dump, analyse, and only then write. Recovery strategy. |
| 1 Nov | Bench kit ordered: SOIC clip, logic analyser, CC3220S LaunchPad, spare flash chips, digital microscope, ESD kit, FPC break-out, ESP32 dev kits. What's spent. |
| 1 Nov | Plan B evaluated: a replacement main board around an ESP32 if the original can't be fixed, a used donor board as a side route. Plan B. |
| 1 Nov | Decision: fix the original board if that can reasonably be done, otherwise build our own. Nothing waits on Anova. |
| 2 Nov | Working assumption: the power board is fine and the problem sits after the 30-pin cable. The test is one supply measurement at the connector. |
| 2 Nov | The central puzzle framed: the Wi-Fi chip's documented rollback mechanism sits on this very board, so a unit that is completely dead after an interrupted update is itself evidence. Three ways to get here. Hypotheses. |
| 2 Nov | Public repository and website live. Dual roadmap published. |
Corrections¶
Statements of substance about the cooker that were changed after publication, with the date. Wording and housekeeping changes aren't listed.
| Date | Correction |
|---|---|
| 2 Nov | "Bricked by the update" was stated as fact. Changed to "died after an interrupted update": the update is the likely cause, not yet a measured one. The measurement that settles it is the first bench job. |