Encoder-decoder models are great for annotation tasks, where you want full information about an input as well as the ability to generate arbitrary structured output. Encoders (BERT) and Decoders (most everything else) only give you one of those.
Google released the open-weights of T5Gemma 2, an encoder-decoder models
Built on top of Gemma 3, T5Gemma 2 is a multimodal, long-context, and heavily multilingual (140 languages) encoder-decoder model, where most models today are decoder-only.
Built on top of Gemma 3, T5Gemma 2 is a multimodal, long-context, and heavily multilingual (140 languages) encoder-decoder model, where most models today are decoder-only.
December 19, 2025 at 3:57 PM
Encoder-decoder models are great for annotation tasks, where you want full information about an input as well as the ability to generate arbitrary structured output. Encoders (BERT) and Decoders (most everything else) only give you one of those.
The Google weather forecast AI use Graph NNs, not Transformer, so in that sense it’s less related to LLMs than most currently deployed ML models are.
On the other hand (as the authors themselves note) GNNs and sparse transformer encoders are different ways of labeling the same operations.
On the other hand (as the authors themselves note) GNNs and sparse transformer encoders are different ways of labeling the same operations.
December 18, 2025 at 5:28 PM
The Google weather forecast AI use Graph NNs, not Transformer, so in that sense it’s less related to LLMs than most currently deployed ML models are.
On the other hand (as the authors themselves note) GNNs and sparse transformer encoders are different ways of labeling the same operations.
On the other hand (as the authors themselves note) GNNs and sparse transformer encoders are different ways of labeling the same operations.
My new vehicle has HDRadio which means I get to hear how misconfigured all my local stations’s encoders are. A few didn’t even sync the latency with FM.
I understand linear media is somewhat of a dying industry, but I do miss when the competition of look & sound was a big deal to station engineers.
I understand linear media is somewhat of a dying industry, but I do miss when the competition of look & sound was a big deal to station engineers.
December 15, 2025 at 3:22 PM
My new vehicle has HDRadio which means I get to hear how misconfigured all my local stations’s encoders are. A few didn’t even sync the latency with FM.
I understand linear media is somewhat of a dying industry, but I do miss when the competition of look & sound was a big deal to station engineers.
I understand linear media is somewhat of a dying industry, but I do miss when the competition of look & sound was a big deal to station engineers.
We've had wonderful success teaching humanists (and other non-CS students) the transformer architecture, incl. things like encoders vs. decoders, stochastic gradient descent, ReLU etc.
It's not the material, it's that most curricula are designed for CS folks and not for a more general audience.
It's not the material, it's that most curricula are designed for CS folks and not for a more general audience.
December 14, 2025 at 4:13 PM
We've had wonderful success teaching humanists (and other non-CS students) the transformer architecture, incl. things like encoders vs. decoders, stochastic gradient descent, ReLU etc.
It's not the material, it's that most curricula are designed for CS folks and not for a more general audience.
It's not the material, it's that most curricula are designed for CS folks and not for a more general audience.
VBR - Variable Bit-Rate. You basically say it how many bits per second you want. It would allow some frames to be bigger and then compensate for other.
Good encoders have "look-ahead" feature that encodes a bunch of frames to get their complexity and distribute the bits proportionally.
Good encoders have "look-ahead" feature that encodes a bunch of frames to get their complexity and distribute the bits proportionally.
December 17, 2025 at 9:52 AM
VBR - Variable Bit-Rate. You basically say it how many bits per second you want. It would allow some frames to be bigger and then compensate for other.
Good encoders have "look-ahead" feature that encodes a bunch of frames to get their complexity and distribute the bits proportionally.
Good encoders have "look-ahead" feature that encodes a bunch of frames to get their complexity and distribute the bits proportionally.
I’ve had an ffmpeg instance running for over 5,000 hours on a x86 machine at work (tv station). More stable than a few of our cheaper hardware encoders, which surprised me.
December 11, 2025 at 1:50 PM
I’ve had an ffmpeg instance running for over 5,000 hours on a x86 machine at work (tv station). More stable than a few of our cheaper hardware encoders, which surprised me.
Like I've got my list of what I desire on the wheel:
- 300mm diameter butterfly/rectangular design
- 2-4 thumb encoders
- 4 or 5 knobs
- 12 face buttons
- 4 paddles on the back, maybe with room for 6
- NO screen
- CHUNKY AF grips
Unsure if I'd want lights for RPM and TC/ABS, but...sounds good?
- 300mm diameter butterfly/rectangular design
- 2-4 thumb encoders
- 4 or 5 knobs
- 12 face buttons
- 4 paddles on the back, maybe with room for 6
- NO screen
- CHUNKY AF grips
Unsure if I'd want lights for RPM and TC/ABS, but...sounds good?
December 13, 2025 at 6:34 PM
Like I've got my list of what I desire on the wheel:
- 300mm diameter butterfly/rectangular design
- 2-4 thumb encoders
- 4 or 5 knobs
- 12 face buttons
- 4 paddles on the back, maybe with room for 6
- NO screen
- CHUNKY AF grips
Unsure if I'd want lights for RPM and TC/ABS, but...sounds good?
- 300mm diameter butterfly/rectangular design
- 2-4 thumb encoders
- 4 or 5 knobs
- 12 face buttons
- 4 paddles on the back, maybe with room for 6
- NO screen
- CHUNKY AF grips
Unsure if I'd want lights for RPM and TC/ABS, but...sounds good?
i was oozing there was some really good motor groan. theeeee uhhhhh rotary encoders I can't remember the name of are my favorite noise. they're like a state machine
December 8, 2025 at 5:02 AM
i was oozing there was some really good motor groan. theeeee uhhhhh rotary encoders I can't remember the name of are my favorite noise. they're like a state machine
Part 3 of MDN’s Image formats series is here! 🖼️
We’ve covered color models & encoders. Now it’s time to compare real codecs and compression tools. Learn how to measure “image quality” the right way.
Read more 👇
developer.mozilla.org/en-US/blog/...
We’ve covered color models & encoders. Now it’s time to compare real codecs and compression tools. Learn how to measure “image quality” the right way.
Read more 👇
developer.mozilla.org/en-US/blog/...
Image formats: Codecs and compression tools | MDN Blog
Image compression involves countless trade-offs between quality, size, and speed. In this final part of the series, we experiment with codecs, metrics, and tools to find practical ways to balance efficiency and visual fidelity.
developer.mozilla.org
December 1, 2025 at 9:00 PM
Part 3 of MDN’s Image formats series is here! 🖼️
We’ve covered color models & encoders. Now it’s time to compare real codecs and compression tools. Learn how to measure “image quality” the right way.
Read more 👇
developer.mozilla.org/en-US/blog/...
We’ve covered color models & encoders. Now it’s time to compare real codecs and compression tools. Learn how to measure “image quality” the right way.
Read more 👇
developer.mozilla.org/en-US/blog/...
The shortest explanation for this is essentially modern encoders, despite theoretically playing nice with any resolution you pipe in, in reality have issues and quality reductions visually from using non 8 divisible resolutions.
So use only divisible by 8 resolutions. 👍
So use only divisible by 8 resolutions. 👍
December 2, 2025 at 3:36 AM
The shortest explanation for this is essentially modern encoders, despite theoretically playing nice with any resolution you pipe in, in reality have issues and quality reductions visually from using non 8 divisible resolutions.
So use only divisible by 8 resolutions. 👍
So use only divisible by 8 resolutions. 👍
🚀 Stop writing hundreds of lines of boilerplate code for Machine Learning in R.
If you’ve used tidymodels or mlr3, you know the workflow can become verbose: recipes, encoders, CV folds, tuning grids, leakage risks…
fastml aims to solve this.
cran.r-project.org/web/packages...
If you’ve used tidymodels or mlr3, you know the workflow can become verbose: recipes, encoders, CV folds, tuning grids, leakage risks…
fastml aims to solve this.
cran.r-project.org/web/packages...
fastml: Fast Machine Learning Model Training and Evaluation
Streamlines the training, evaluation, and comparison of multiple machine learning models with minimal code by providing comprehensive data preprocessing and support for a wide range of algorithms with...
cran.r-project.org
November 28, 2025 at 11:29 AM
🚀 Stop writing hundreds of lines of boilerplate code for Machine Learning in R.
If you’ve used tidymodels or mlr3, you know the workflow can become verbose: recipes, encoders, CV folds, tuning grids, leakage risks…
fastml aims to solve this.
cran.r-project.org/web/packages...
If you’ve used tidymodels or mlr3, you know the workflow can become verbose: recipes, encoders, CV folds, tuning grids, leakage risks…
fastml aims to solve this.
cran.r-project.org/web/packages...
Encoderfile v0.1.0 is here.
A new way to deploy encoders as single-binary executables with no runtime dependencies. Faster, leaner, and built for correctness.
Full details here: https://link.mozilla.ai/encoderfile-v010
A new way to deploy encoders as single-binary executables with no runtime dependencies. Faster, leaner, and built for correctness.
Full details here: https://link.mozilla.ai/encoderfile-v010
November 24, 2025 at 5:16 PM
Encoderfile v0.1.0 is here.
A new way to deploy encoders as single-binary executables with no runtime dependencies. Faster, leaner, and built for correctness.
Full details here: https://link.mozilla.ai/encoderfile-v010
A new way to deploy encoders as single-binary executables with no runtime dependencies. Faster, leaner, and built for correctness.
Full details here: https://link.mozilla.ai/encoderfile-v010
Finished building my custom keyboard! I'm so darn happy with this dark wood grain look.
Right side has a trackball for mouse, and the left side has a touchpad for scrolling. Both sides have rotary encoders that I can bind to various actions.
Right side has a trackball for mouse, and the left side has a touchpad for scrolling. Both sides have rotary encoders that I can bind to various actions.
November 20, 2025 at 8:48 AM
Finished building my custom keyboard! I'm so darn happy with this dark wood grain look.
Right side has a trackball for mouse, and the left side has a touchpad for scrolling. Both sides have rotary encoders that I can bind to various actions.
Right side has a trackball for mouse, and the left side has a touchpad for scrolling. Both sides have rotary encoders that I can bind to various actions.
Part 3 of MDN’s Image formats series is here! 🖼️
We’ve covered color models & encoders. Now it’s time to compare real codecs and compression tools. Learn how to measure “image quality” the right way.
Read more 👇
developer.mozilla.org/en-US/blog/...
We’ve covered color models & encoders. Now it’s time to compare real codecs and compression tools. Learn how to measure “image quality” the right way.
Read more 👇
developer.mozilla.org/en-US/blog/...
Image formats: Codecs and compression tools | MDN Blog
Image compression involves countless trade-offs between quality, size, and speed. In this final part of the series, we experiment with codecs, metrics, and tools to find practical ways to balance efficiency and visual fidelity.
developer.mozilla.org
November 18, 2025 at 4:00 PM
Part 3 of MDN’s Image formats series is here! 🖼️
We’ve covered color models & encoders. Now it’s time to compare real codecs and compression tools. Learn how to measure “image quality” the right way.
Read more 👇
developer.mozilla.org/en-US/blog/...
We’ve covered color models & encoders. Now it’s time to compare real codecs and compression tools. Learn how to measure “image quality” the right way.
Read more 👇
developer.mozilla.org/en-US/blog/...
I still need to build my klor, got stalled halfway through. I may do that today.
November 15, 2025 at 5:19 PM
I still need to build my klor, got stalled halfway through. I may do that today.
Finally got the encoders in my studio to be stable, thanks to Genlock. Still some stuff to tweak, but on the right track for doing a show next week. Gonna snuggle up with a partner in another city this weekend and dabble in writing some documentation in the mean time. ^.^
November 13, 2025 at 4:48 AM
Finally got the encoders in my studio to be stable, thanks to Genlock. Still some stuff to tweak, but on the right track for doing a show next week. Gonna snuggle up with a partner in another city this weekend and dabble in writing some documentation in the mean time. ^.^
Got a little MIDI/DAW controller to play around with and… uh… seriously?
November 12, 2025 at 6:59 PM
Got a little MIDI/DAW controller to play around with and… uh… seriously?
For the next project, I'm trying to make sure the color balance is accurate, which is how I noticed the rendering issue. After messing with the rendering output settings and the encoders, I think I may have fixed it 💡
November 7, 2025 at 4:10 AM
For the next project, I'm trying to make sure the color balance is accurate, which is how I noticed the rendering issue. After messing with the rendering output settings and the encoders, I think I may have fixed it 💡
I bought an iMac Studio for editing when my last PC died. But there is a weird quirk with Macs with certain rendering software with how QuickTime encoders handle colors differently
November 7, 2025 at 4:08 AM
I bought an iMac Studio for editing when my last PC died. But there is a weird quirk with Macs with certain rendering software with how QuickTime encoders handle colors differently
Are you talking about custom designing and sourcing a pcb, switches, rotary encoders, keycaps, a case, and then programming a microcontroller? Or looking for suggestions for a mostly prebuilt keyboard with those features ?
November 5, 2025 at 1:07 AM
Are you talking about custom designing and sourcing a pcb, switches, rotary encoders, keycaps, a case, and then programming a microcontroller? Or looking for suggestions for a mostly prebuilt keyboard with those features ?
OK this thing is actually really fascinating, by @efog.tech. Higher end spec than the Nano/Adept. Comes with a BTU option if you want. Shipping is apparently rough but it's open source. Dual encoders. BLUETOOTH. DIY/semi-DIY kit only. Might have to make this or get some PCBs printed.
November 2, 2025 at 6:38 PM
OK this thing is actually really fascinating, by @efog.tech. Higher end spec than the Nano/Adept. Comes with a BTU option if you want. Shipping is apparently rough but it's open source. Dual encoders. BLUETOOTH. DIY/semi-DIY kit only. Might have to make this or get some PCBs printed.
i forget which encoding speed preset i used for the video encoders but it could always be something slower that makes better use of the bits. also the HLS streams are all CBR so it can't do any kind of VBR stuff to use bits where they're more important for retaining visual fidelity
November 3, 2025 at 1:54 AM
i forget which encoding speed preset i used for the video encoders but it could always be something slower that makes better use of the bits. also the HLS streams are all CBR so it can't do any kind of VBR stuff to use bits where they're more important for retaining visual fidelity
PII detection for 15x-500x cheaper
Goodfire and Rakuten used sparse auto encoders (SAEs, a mech interp thing) to detect PII
an SAE is a secondary model trained on the primary LLM, they used a random forest. This 2nd model tells you what “features” were activated
www.goodfire.ai/research/rak...
Goodfire and Rakuten used sparse auto encoders (SAEs, a mech interp thing) to detect PII
an SAE is a secondary model trained on the primary LLM, they used a random forest. This 2nd model tells you what “features” were activated
www.goodfire.ai/research/rak...
Deploying Interpretability to Production with Rakuten: SAE Probes for PII Detection
www.goodfire.ai
October 30, 2025 at 11:51 AM
PII detection for 15x-500x cheaper
Goodfire and Rakuten used sparse auto encoders (SAEs, a mech interp thing) to detect PII
an SAE is a secondary model trained on the primary LLM, they used a random forest. This 2nd model tells you what “features” were activated
www.goodfire.ai/research/rak...
Goodfire and Rakuten used sparse auto encoders (SAEs, a mech interp thing) to detect PII
an SAE is a secondary model trained on the primary LLM, they used a random forest. This 2nd model tells you what “features” were activated
www.goodfire.ai/research/rak...
The reason why frogs are so effective at resisting the Trump regime, is because frogs is a latent (in neurology it means a deep signal in your brain and in machine learning it's the the middle layers in auto-encoders) representation of your own gut reaction, which does not believe propaganda.
🐸🐸🐸
🐸🐸🐸
October 22, 2025 at 1:07 AM
The reason why frogs are so effective at resisting the Trump regime, is because frogs is a latent (in neurology it means a deep signal in your brain and in machine learning it's the the middle layers in auto-encoders) representation of your own gut reaction, which does not believe propaganda.
🐸🐸🐸
🐸🐸🐸
fwiw my mental model (not great) is that encoders are great at compressing complex inputs (specialized, as you say), decoders are general, and encoder-decoders are the best of both. i.e. you don’t have to go through the encoder half on every single output token
October 21, 2025 at 1:42 PM
fwiw my mental model (not great) is that encoders are great at compressing complex inputs (specialized, as you say), decoders are general, and encoder-decoders are the best of both. i.e. you don’t have to go through the encoder half on every single output token