M
Team M37Labs
knowledge-base
Announcing Web Search on Amazon Bedrock AgentCore: Ground your AI agents in current…
Link
Jun 19, 2026 • 5:00 AMM37Labs Feed
Knowledge base
Direct from our engineering and research Slack channels. Raw insights, technical breakdowns, field notes, and real-time updates from the M37Labs team.
M
Team M37Labs
knowledge-base
has anyone tried this
has anyone tried this https://youtu.be/OvLIae4HCeM?si=FqQyS4XPLOAU3Umj
Link
Jun 18, 2026 • 9:51 AMM
Team M37Labs
knowledge-base
DiffusionGemma: 4x faster text generation
Link
Jun 12, 2026 • 5:01 AMM
Team M37Labs
knowledge-base
This part is :explodinghead: :
https://claude.com/blog/introducing-dynamic-workflows-in-claude-code
This part is :exploding_head: :
This part is :exploding_head: :
Code Snippet
Rewriting Bun with dynamic workflows :
An example of what dynamic workflows can unlock at scale is the recent rewrite of Bun. Jarred Sumner used dynamic workflows to port Bun from Zig to Rust with 99.8% of the existing test suite passing, roughly 750,000 lines of Rust, and eleven days from first commit to merge. One workflow mapped the right Rust lifetime for every struct field in the Zig codebase. The next wrote every .rs file as a behavior-identical port of its .zig counterpart, hundreds of agents working in parallel with two reviewers on each file. A fix loop then drove the build and test suite until both ran clean. After the port landed, an overnight workflow addressed unnecessary data copies and opened a PR for each for final review. While not yet in production, all of this was handled by dynamic workflows. Jarred will be writing about this more in the future.
LinkCode
Jun 11, 2026 • 6:50 PMM
Team M37Labs
knowledge-base
The AI Dev cycle
The AI Dev cycle
Expand Image
Image
Jun 6, 2026 • 7:18 AMM
Team M37Labs
knowledge-base
For summarised in depth understanding of AI Models - made by practitioners, YC paper club…
https://www.youtube.com/watch?v=wE1ZgJdt4uM
For summarised in depth understanding of AI Models - made by practitioners, YC paper club has some fantastic videos,
For summarised in depth understanding of AI Models - made by practitioners, YC paper club has some fantastic videos,
Link
Jun 1, 2026 • 5:13 AMM
Team M37Labs
knowledge-base
An OpenAI model has disproved a central conjecture in discrete geometry
Link
May 25, 2026 • 4:40 AMM
Team M37Labs
knowledge-base
#WeekendHackathon, Updated 7B parameter to 14B parameters. And it worked like a charm on…
#WeekendHackathon, Updated 7B parameter to 14B parameters. And it worked like a charm on M3 Pro. LM studio updates are crazy. Next step is to learn proper finetuning. Idea is to have LLM running locally, and use Graphql for managing memory context.
Expand Image
Image
May 24, 2026 • 12:44 PMM
Team M37Labs
knowledge-base
Everything Announced at Google I/O 2026
Link
May 21, 2026 • 1:40 AMM
Team M37Labs
knowledge-base
All the news from the Google I/O 2026 Developer keynote- Google Developers Blog
Link
May 21, 2026 • 1:39 AMM
Team M37Labs
knowledge-base
tinyhumansai/openhuman: Your Personal AI super intelligence. Private, Simple and…
tinyhumansai/openhuman: Your Personal AI super intelligence. Private, Simple and extremely powerful. https://share.google/dnxSd9FSrGmPLiO4F
Link
May 18, 2026 • 4:51 PMM
Team M37Labs
knowledge-base
Introducing Claude for Small Business
Link
May 14, 2026 • 3:04 AMM
Team M37Labs
knowledge-base
Accelerating Gemma 4: faster inference with multi-token prediction drafters
Link
May 11, 2026 • 7:33 PMM
Team M37Labs
knowledge-base
I am keen to dive deep into this space, slowly inching forward.
I am keen to dive deep into this space, slowly inching forward.
Expand Image
Image
May 11, 2026 • 7:23 PMM
Team M37Labs
knowledge-base
Verbalized Sampling: How to Mitigate Mode Collapse and Unlock LLM Diversity
Verbalized Sampling: How to Mitigate Mode Collapse and Unlock LLM Diversity
Jiayi Zhang, Simon Yu, Derek Chong, Anthony Sicilia, Michael R. Tomz, Christopher D. Manning, Weiyan Shi
> Post-training alignment often reduces LLM diversity, leading to a phenomenon known as mode collapse. Unlike prior work that attributes this effect to algorithmic limitations, we identify a fundamental, pervasive data-level driver: typicality bias in preference data, whereby annotators systematically favor familiar text as a result of well-established findings in cognitive psychology. We formalize this bias theoretically, verify it on preference datasets empirically, and show that it plays a central role in mode collapse. Motivated by this analysis, we introduce Verbalized Sampling, a simple, training-free prompting strategy to circumvent mode collapse. VS prompts the model to verbalize a probability distribution over a set of responses (e.g., “Generate 5 jokes about coffee and their corresponding probabilities”). Comprehensive experiments show that VS significantly improves performance across creative writing (poems, stories, jokes), dialogue simulation, open-ended QA, and synthetic data generation, without sacrificing factual accuracy and safety. For instance, in creative writing, VS increases diversity by 1.6-2.1x over direct prompting. We further observe an emergent trend that more capable models benefit more from VS. In sum, our work provides a new data-centric perspective on mode collapse and a practical inference-time remedy that helps unlock pre-trained generative diversity.
Jiayi Zhang, Simon Yu, Derek Chong, Anthony Sicilia, Michael R. Tomz, Christopher D. Manning, Weiyan Shi
> Post-training alignment often reduces LLM diversity, leading to a phenomenon known as mode collapse. Unlike prior work that attributes this effect to algorithmic limitations, we identify a fundamental, pervasive data-level driver: typicality bias in preference data, whereby annotators systematically favor familiar text as a result of well-established findings in cognitive psychology. We formalize this bias theoretically, verify it on preference datasets empirically, and show that it plays a central role in mode collapse. Motivated by this analysis, we introduce Verbalized Sampling, a simple, training-free prompting strategy to circumvent mode collapse. VS prompts the model to verbalize a probability distribution over a set of responses (e.g., “Generate 5 jokes about coffee and their corresponding probabilities”). Comprehensive experiments show that VS significantly improves performance across creative writing (poems, stories, jokes), dialogue simulation, open-ended QA, and synthetic data generation, without sacrificing factual accuracy and safety. For instance, in creative writing, VS increases diversity by 1.6-2.1x over direct prompting. We further observe an emergent trend that more capable models benefit more from VS. In sum, our work provides a new data-centric perspective on mode collapse and a practical inference-time remedy that helps unlock pre-trained generative diversity.
Apr 28, 2026 • 12:38 AM