I noticed a strange things.
I have, now, over 4200 points with 37 wus correctly validated, so approximately 110 points for a single wu.
But if i go to my profiles every single wu is between 3 and 5 points...

There must be some kind of streaming credit still, beyond the credit given by the task.

With the new app the situation seems normalized...

Zero credit on valid tasks on the past several app versions. From I think 3.47 to 3.52 all junk.

Edit: And 3.53

Up to v3.57 with no credit.

Now there are some WUs who receive 100 cr, some 9,73 and other 4,86 for the same amount of working seconds
Is this correct ??

All mine seem to have been adjusted to 100 per task so I lost 75% of my credit while others are not getting 1.7mil per task.

Now it's 200 credits per WU

From 200 per task to 0 to 2 credit per task. Mostly around 0.1

Now 10.75 credits/WU

In reply to mmonnin's message of 8 Feb 2026:

From 200 per task to 0 to 2 credit per task. Mostly around 0.1

I noticed this in my zero-credit-wus

[Trainer] Sync interval: 999999999 samples
[WeightSync] Connecting to ws://65.21.196.61:8765...
[WeightSync] Connection error: timed out during opening handshake
[WeightSync] Connecting to ws://65.21.196.61:8765...
[WeightSync] Connection error: timed out during opening handshake
[WeightSync] Connecting to ws://65.21.196.61:8765...
[BOINC] Training stopped. 256 total samples
[BOINC] Self-testing on 2048 bytes holdout...
[BOINC] Self-test: bpc=0.0256 acc=100.0% (63 samples)
[WeightSync] Connection error: timed out during opening handshake

But my pc is correctly connected to internet....

I do not see that. Back to back tasks

200 credit:
<core_client_version>8.2.5</core_client_version>
<![CDATA[
<stderr_txt>
2026-02-08 09:25:33 (27176): BOINC streaming wrapper starting
2026-02-08 09:25:33 (27176): Streaming wrapper: running ../../projects/axiom.heliex.net/axiom_worker_3.57_x86_64-pc-linux-gnu__cuda (wu.json result.json)
2026-02-08 09:25:33 (27176): Streaming wrapper: created child process 27178
[GPU] CuPy CUDA: 15839MB total, 11946MB free
[BOINC] Starting BOINC mode
[BOINC] GPU detected: 15839MB VRAM, 11946MB free
[BOINC] Work unit: wu.json
[BOINC] Result file: result.json
[BOINC] Current directory: /var/lib/boinc10/slots/0
[BOINC] Directory contents: ['job.xml', 'result.json', 'boinc_mmap_file', 'init_data.xml', 'boinc_lockfile', 'stderr.txt', 'boinc_setup_complete', 'wu.json', 'axiom_worker_3.57_x86_64-pc-linux-gnu__cuda', 'streaming_wrapper_checkpoint.txt', 'streaming_wrapper_300_x86_64-pc-linux-gnu']
[BOINC] Resolved soft link: ../../projects/axiom.heliex.net/wu_9_g493263_62857_input.json
[BOINC] Resolved soft link: ../../projects/axiom.heliex.net/wu_9_g493263_62857_0_r844329907_0
[BOINC] Actual WU file: ../../projects/axiom.heliex.net/wu_9_g493263_62857_input.json
[BOINC] Successfully loaded work unit
[BOINC] Expert: 9
[BOINC] Run duration: 150
[BOINC] Sync interval: 999999999
[Trainer] Initializing expert 9...
[BOINC] Error: CUDA_ERROR_NO_BINARY_FOR_GPU: no kernel image is available for execution on the device
[BOINC] Complete: syncs=0, samples=0
2026-02-08 09:25:47 (27176): axiom_worker_3.57_x86_64-pc-linux-gnu__cuda exited; CPU time 7.461367
2026-02-08 09:25:47 (27176): called boinc_finish(0)

</stderr_txt>
]]>

0 Credit:
<core_client_version>8.2.5</core_client_version>
<![CDATA[
<stderr_txt>
2026-02-08 09:27:27 (27240): BOINC streaming wrapper starting
2026-02-08 09:27:27 (27240): Streaming wrapper: running ../../projects/axiom.heliex.net/axiom_worker_3.57_x86_64-pc-linux-gnu__cuda (wu.json result.json)
2026-02-08 09:27:27 (27240): Streaming wrapper: created child process 27242
[GPU] CuPy CUDA: 15839MB total, 11946MB free
[BOINC] Starting BOINC mode
[BOINC] GPU detected: 15839MB VRAM, 11946MB free
[BOINC] Work unit: wu.json
[BOINC] Result file: result.json
[BOINC] Current directory: /var/lib/boinc10/slots/0
[BOINC] Directory contents: ['job.xml', 'result.json', 'boinc_mmap_file', 'init_data.xml', 'boinc_lockfile', 'stderr.txt', 'boinc_setup_complete', 'wu.json', 'axiom_worker_3.57_x86_64-pc-linux-gnu__cuda', 'streaming_wrapper_checkpoint.txt', 'streaming_wrapper_300_x86_64-pc-linux-gnu']
[BOINC] Resolved soft link: ../../projects/axiom.heliex.net/wu_9_g493264_5215_input.json
[BOINC] Resolved soft link: ../../projects/axiom.heliex.net/wu_9_g493264_5215_0_r1834708217_0
[BOINC] Actual WU file: ../../projects/axiom.heliex.net/wu_9_g493264_5215_input.json
[BOINC] Successfully loaded work unit
[BOINC] Expert: 9
[BOINC] Run duration: 150
[BOINC] Sync interval: 999999999
[Trainer] Initializing expert 9...
[BOINC] Error: CUDA_ERROR_NO_BINARY_FOR_GPU: no kernel image is available for execution on the device
[BOINC] Complete: syncs=0, samples=0
2026-02-08 09:27:40 (27240): axiom_worker_3.57_x86_64-pc-linux-gnu__cuda exited; CPU time 7.580233
2026-02-08 09:27:40 (27240): called boinc_finish(0)

</stderr_txt>
]]>

I have highlighted the differences. It's all on the server side.