Replies: 2 comments
|
Summary and transcript are now in the top post. |
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
Summary generated by AI from the meeting recording.
Post-Meeting Summary and Transcript
Attendees: Ryan (host), Aadhi, Dylan, Econoalchemist, Johnny, Schnitzel, Skot (joined late), Tyler.
Referenced but absent: Jayr, Ronald, Sean (ixtech.xyz), Average Gary.
Summary
Opening and a new agenda order (0:00)
Econoalchemist, the 256 Foundation's project manager, joined his first dev call; his schedule now has Mondays reserved for the Foundation, so he plans to be a regular. Ryan explained this call's inverted agenda: earlier calls followed an item's path through the project (chat, then discussion, then issue, then PR), which spent the call's best time on the most nebulous topics. This call runs that order backward, PRs first, with a new section at the top for announcements and host topics.
Coldcard PSA and user promises (4:58)
After a PSA on the Coldcard entropy bug, Ryan drew a lesson for the project: a short list of promises Mujina owes its users, in rough order of importance. Don't burn someone's house or business down. Don't destroy someone's equipment. Don't waste someone's electricity. Protect hashrate from being stolen. Protect dev machines from compromises. On hashrate theft, mining to a pool gives users an external cross-check that shares go where they intend, an architectural argument for keeping template creation out of Mujina. Schnitzel described resale attacks seeded by unprotected UIs and default root passwords, and asked whether Mujina's API will ship with a password by default; Ryan intends authentication and permissions soon and is still thinking about the out-of-the-box posture, possibly key provisioning at image-load time in the style of the Raspberry Pi loader. Johnny added that much of the protection story belongs to the OS image (firewalls, secure boot) and suggested auditing the Stratum client against malicious pools. On supply chain risk, Ryan wants a Cargo lockfile immediately, plus an update rule that holds back crate versions younger than about a week. The promises list may become a GitHub discussion.
Website refresh (16:35)
Prompted by the last call's compatibility matrix discussion, Ryan refreshed mujina.org as a landing page and the future home of user documentation, including a hardware compatibility page that already links a couple of contributor forks. The site builds from its own repo, 256foundation/mujina-website, and PRs are welcome.
devpool (19:24)
Ryan packaged the classroom Bitcoin network from his Nairobi mining workshop as devpool, a public repo on his account: a container stack with a modified Hydrapool, bitcoind on signet, and dashboards, runnable under Docker or Podman from prebuilt images. Fractional difficulty and fast retargeting let anything from a CPU miner to an S19 register on the dashboards, which public pools cannot show for CPU miners.
A bm13xx chip reference (21:09)
Ryan merged the growth of the Bitmain chip protocol doc into a reference of a few thousand lines, collecting what he has learned building his unified Bitmain chip driver: material from the web, ESP-Miner, his own reverse engineering, and captures of factory and third-party firmware. He pitched it as a missing datasheet, readable enough to teach a newcomer how these chips work, and as an on-ramp for contributors: the document marks what is fuzzy or unexplored, and running experiments against those parts and folding the results back in is a worthwhile contribution in itself. He will share it with adjacent projects for review. On Schnitzel's suggestion, Ryan will link it from mujina.org so people and agents can find it. Aadhi offered a matching reference doc for the Canaan 3197S chip series, using this one as the template.
BZM2 progress and the Bitaxe Bonanza (26:25)
Ronald has pushed several BZM2 PRs that Ryan has not yet reviewed closely; his general chip documentation now lives in its own repo under his Blockscale Solutions account, which resolves the earlier concern about reference docs in the Mujina tree. Johnny reported the Bitaxe Bonanza is now fully functional as a USB miner (one DisplayPort fix left for Skot), so he returns to Mujina this week: updating his fork and PR, folding in Ronald's work where it fits, and possibly writing a BZM2 reference in the style of Ryan's. A second device is in progress in parallel. Ryan has no BZM2 hardware; Johnny will ask Skot about that.
Contribution advice: vertical slices (31:05)
Prompted by Ronald's type changes and a configuration conversation with Jayr, Ryan asked that contributions arrive in the context of making something concrete work, not as isolated types or abstract buildup toward a big plan. Pick one feature, like setting the pool URL over the API, and build the narrowest vertical slice through the stack that delivers it, coherent with the larger design. A slice can be run, tested, reviewed, and merged; a type modified for the benefit of an unmerged fork cannot. First versions of genuinely new hardware support are the exception, since they need critical mass, but the history should still arrive in logical chunks.
A tracking issue for pool configuration (34:49)
Jayr will start implementing MIP-0001 with pool configuration as the first concrete slice. Ryan filed a tracking issue so related issues, PRs, and commits have a handle to reference, and offered the same to anyone with long-running work.
Direct-to-node mining, and what Mujina is not (36:28)
The only discussion activity was on Ryan's direct-to-node experiment, and it built the rationale for keeping job creation outside Mujina: template creation is a hard problem in its own right, it runs most efficiently next to the node, multiple Mujinas should not each build their own jobs, and an external job source cross-checks that Mujina's shares are real work. Schnitzel suggested mujina.org state plainly what Mujina is not, since AI-speed contribution increasingly adds code to projects where it does not belong; Ryan and Johnny endorsed the idea.
Heatpunk Summit 2027 (40:24)
The Heatpunk Summit announced its dates, February 26 and 27 in Denver, and opened a waitlist that already holds 50 to 60 names. Tyler is planning 256 Foundation demos and workshops. Ryan recommends the summit as high signal and directly relevant, and hopes most contributors can attend.
Power policies (42:08)
From the Telegram chat: a new user's questions prompted a thread on what Mujina would call scheduling policies, whether Mujina aims for maximum efficiency or a power target, and how power targets would be distributed across boards. Ryan restated the vision the project has already discussed; nothing new was invented.
The APW12's hidden current sensor (43:11)
Skot thinks the APW12's internal PIC microcontroller may already measure current: current feedback is central to voltage regulation, the PIC has ADCs, and he recalls the schematic (which the community has) hooking a current measurement to it. The ways in: an ICSP header sits next to the PIC, so the firmware can be dumped if readback security is off and reprogrammed with a PICkit 3 either way, and he has an APW3 firmware dump ready for Ghidra. Schnitzel has a financial incentive to see this happen for the Nova project. Real current readings from the PSU would feed tuning, and the safety monitoring Ryan wants for temperatures and currents.
Community hardware: Nano 3S, BMM101, NerdQAxe++ (49:25)
Aadhi's Avalon Nano 3S is hashing under Mujina. He mirrored Canaan's own split on the K230D processor: the Linux core runs Mujina and handles Stratum and power control, the RISC-V RTOS core talks to the ASIC (preambles, CRC, telemetry), and the two sides share jobs over Canaan's IPC mechanism. He will clean up the code and open source it. On the BMM101, the Braiins CEO shared control board and hash board documentation; that port needs more time. Ryan noted Sean has Mujina running on a NerdQAxe++ in his fork.
Open floor: mining by hand (53:32)
Schnitzel computed a human at 33 microhashes per second and threatened a PR for the Homo Sapiens 1 chip. The group specified telemetry (temperature, glucose in millimole per liter, live over Bluetooth from a glucose patch) and a dashboard with a form for hand-calculating hashes. Ryan declared it a good April Fools Day PR.
Action items
Transcript
Full transcript on the wiki: https://github.com/256foundation/mujina/wiki/dev-call-20260803
Agenda (as planned)
Agenda
Note: These meetings are recorded and transcribed by AI.
Purpose: touch news and happenings to keep everyone aware of what's happening,
roughly follow development workflow, not a replacement for the other communication
channels
Announcements & host topics
Coldcard security risk: immediate action required
be brute forced.
documentation, built from
256foundation/mujina-website.
An early start, incomplete, and PRs are welcome.
signet, Hydrapool, the mempool explorer, and dashboards.
Fractional difficulty lets a laptop CPU hold a session, and the
pool paces block submission to about 30 seconds so the explorer
stays readable. Published images for amd64 and arm64, so a clone
and
just upis the whole setup.PRs
Running experiments and improving the research in a
document like this is a worthwhile activity and
contribution.
bzm2-hwref,
resolving the in-tree docs concern from the June 15 call.
Issues
out of review discussion on the MIP PR,
MIP-0001: Configuration and API mujina-mips#1.
Discussions
simply as possible from a local node" use case may be best
served by a separate upstream program that speaks
getblocktemplate, not by the miner itself.
Forum & Chat
Mujina, Libre Board, Ember One, and Hydrapool.
source design that mixes solar and wall power and would tell
Mujina how much it can draw
(thread).
real current readings
(thread), and separately
a smart bus bar module for current measurement
(thread).
Canaan A3197 protocol, which bears on the Avalon Nano 3s
discussion, Support for Avalon Nano 3s #25
(message).
adding NerdQAxe++ boards to his test setup
(message).
Follow-up
Other
Open floor
All reactions