Andrei Onel
@onel.bsky.social
Founder askmanu, helping companies document their code base
Pro-active mode is the future of AI agents
At the end of November we had the AI Engineer Code summit happening in New York with a packed list of top-notch speakers.
One talk that really stood out was Kat Korevec’s from the Google Labs team.
Link here 👉 www.youtube.com/live/cMSprbJ...
At the end of November we had the AI Engineer Code summit happening in New York with a packed list of top-notch speakers.
One talk that really stood out was Kat Korevec’s from the Google Labs team.
Link here 👉 www.youtube.com/live/cMSprbJ...
AIE CODE 2025: AI Leadership ft Anthropic, OpenAI, McKinsey, Bloomberg, Google Deepmind, and Tenex
YouTube video by AI Engineer
www.youtube.com
December 11, 2025 at 2:12 PM
Pro-active mode is the future of AI agents
At the end of November we had the AI Engineer Code summit happening in New York with a packed list of top-notch speakers.
One talk that really stood out was Kat Korevec’s from the Google Labs team.
Link here 👉 www.youtube.com/live/cMSprbJ...
At the end of November we had the AI Engineer Code summit happening in New York with a packed list of top-notch speakers.
One talk that really stood out was Kat Korevec’s from the Google Labs team.
Link here 👉 www.youtube.com/live/cMSprbJ...
Reposted by Andrei Onel
It's finally here: The public (and most complete) version of my talk covering every stage of the process to build Olmo 3 Think. This involves changes and new considerations of every angle of the stack, from pretraining, evaluation, and of course post-training.
December 10, 2025 at 7:36 PM
It's finally here: The public (and most complete) version of my talk covering every stage of the process to build Olmo 3 Think. This involves changes and new considerations of every angle of the stack, from pretraining, evaluation, and of course post-training.
Reposted by Andrei Onel
It's ChatGPT's third birthday today, so I pulled other some quotes from the early
months of ChatGPT on my blog
OpenAI really did not expect it to be as popular as it was!
simonwillison.net/2025/Nov/30/...
months of ChatGPT on my blog
OpenAI really did not expect it to be as popular as it was!
simonwillison.net/2025/Nov/30/...
ChatGPT is three years old today
It's ChatGPT's third birthday today. It's fun looking back at Sam Altman's low key announcement thread from November 30th 2022: today we launched ChatGPT. try talking with it here: chat.openai.com …
simonwillison.net
November 30, 2025 at 10:21 PM
It's ChatGPT's third birthday today, so I pulled other some quotes from the early
months of ChatGPT on my blog
OpenAI really did not expect it to be as popular as it was!
simonwillison.net/2025/Nov/30/...
months of ChatGPT on my blog
OpenAI really did not expect it to be as popular as it was!
simonwillison.net/2025/Nov/30/...
Europe is softening GDPR the right way 🇪🇺 and a privacy PSA
I’m a fan of @theverge.com but I think this recent article didn’t do a good job at explaining the new changes proposed to the GDPR.
www.theverge.com/news/823750/...
I’m a fan of @theverge.com but I think this recent article didn’t do a good job at explaining the new changes proposed to the GDPR.
www.theverge.com/news/823750/...
Europe is scaling back its landmark privacy and AI laws
The EU folds under Big Tech’s pressure.
www.theverge.com
November 20, 2025 at 1:11 PM
Europe is softening GDPR the right way 🇪🇺 and a privacy PSA
I’m a fan of @theverge.com but I think this recent article didn’t do a good job at explaining the new changes proposed to the GDPR.
www.theverge.com/news/823750/...
I’m a fan of @theverge.com but I think this recent article didn’t do a good job at explaining the new changes proposed to the GDPR.
www.theverge.com/news/823750/...
Reminder: AI Collective is launching in Dublin tomorrow
If you're an AI enthusiast join us for drinks and meet some amazing people.
🌟 We have some great things planned.
luma.com/startdubchap...
If you're an AI enthusiast join us for drinks and meet some amazing people.
🌟 We have some great things planned.
luma.com/startdubchap...
🧠 The AI Collective is setting foot in Dublin!🍀 · Luma
The AI Collective is landing in one of Europe’s most exciting tech cities — Dublin! We’re here to bring together Ireland’s AI thinkers, makers, and doers to: ✨…
luma.com
November 18, 2025 at 4:52 PM
Reminder: AI Collective is launching in Dublin tomorrow
If you're an AI enthusiast join us for drinks and meet some amazing people.
🌟 We have some great things planned.
luma.com/startdubchap...
If you're an AI enthusiast join us for drinks and meet some amazing people.
🌟 We have some great things planned.
luma.com/startdubchap...
🚀 The big new feature that I'm really excited in is live: you can now create a documentation issue and @askmanu.bsky.social will go ahead and create the perfect documentation for your project.
𝗬𝗼𝘂 𝗰𝗮𝗻 𝗻𝗼𝘄 𝗮𝘀𝘀𝗶𝗴𝗻 𝗚𝗶𝘁𝗵𝘂𝗯 𝗶𝘀𝘀𝘂𝗲𝘀 𝘁𝗼 𝗮𝘀𝗸𝗺𝗮𝗻𝘂 🌟
The biggest power of askmanu is that it pro-actively looks for things to document in your repo: a comprehensive readme, great architecture overview, step by step dev setup guide, etc.
The biggest power of askmanu is that it pro-actively looks for things to document in your repo: a comprehensive readme, great architecture overview, step by step dev setup guide, etc.
November 5, 2025 at 5:09 PM
🚀 The big new feature that I'm really excited in is live: you can now create a documentation issue and @askmanu.bsky.social will go ahead and create the perfect documentation for your project.
Reposted by Andrei Onel
Training LLMs end to end is hard. But way more people should, and will, be doing it in the future.
The @hf.co Research team is excited to share their new e-book that covers the full pipeline:
· pre-training,
· post-training,
· infra.
200+ pages of what worked and what didn’t. ⤵️
The @hf.co Research team is excited to share their new e-book that covers the full pipeline:
· pre-training,
· post-training,
· infra.
200+ pages of what worked and what didn’t. ⤵️
November 2, 2025 at 3:17 PM
Training LLMs end to end is hard. But way more people should, and will, be doing it in the future.
The @hf.co Research team is excited to share their new e-book that covers the full pipeline:
· pre-training,
· post-training,
· infra.
200+ pages of what worked and what didn’t. ⤵️
The @hf.co Research team is excited to share their new e-book that covers the full pipeline:
· pre-training,
· post-training,
· infra.
200+ pages of what worked and what didn’t. ⤵️
Reposted by Andrei Onel
We just shipped again 🚢 you can now 𝗮𝘀𝗸 𝗺𝗮𝗻𝘂 𝘁𝗼 𝗺𝗮𝗸𝗲 𝘂𝗽𝗱𝗮𝘁𝗲𝘀 𝘁𝗼 𝗣𝗥𝘀.
The biggest feature request by far was for a way to ask for updates to a PR.
You can now @ at manu and it will make updates to the created documents
The biggest feature request by far was for a way to ask for updates to a PR.
You can now @ at manu and it will make updates to the created documents
October 28, 2025 at 1:11 PM
We just shipped again 🚢 you can now 𝗮𝘀𝗸 𝗺𝗮𝗻𝘂 𝘁𝗼 𝗺𝗮𝗸𝗲 𝘂𝗽𝗱𝗮𝘁𝗲𝘀 𝘁𝗼 𝗣𝗥𝘀.
The biggest feature request by far was for a way to ask for updates to a PR.
You can now @ at manu and it will make updates to the created documents
The biggest feature request by far was for a way to ask for updates to a PR.
You can now @ at manu and it will make updates to the created documents
Reposted by Andrei Onel
One of the cool open source projects using askmanu for their docs is 5ire.
A beautifully built AI assistant client and MCP client.
Check out the project here:
github.com/nanbingxyz/5...
A beautifully built AI assistant client and MCP client.
Check out the project here:
github.com/nanbingxyz/5...
GitHub - nanbingxyz/5ire: 5ire is a cross-platform desktop AI assistant, MCP client. It compatible with major service providers, supports local knowledge base and tools via model context protocol se...
5ire is a cross-platform desktop AI assistant, MCP client. It compatible with major service providers, supports local knowledge base and tools via model context protocol servers . - nanbingxyz/5ire
github.com
September 16, 2025 at 2:54 PM
One of the cool open source projects using askmanu for their docs is 5ire.
A beautifully built AI assistant client and MCP client.
Check out the project here:
github.com/nanbingxyz/5...
A beautifully built AI assistant client and MCP client.
Check out the project here:
github.com/nanbingxyz/5...
Super happy to announce that peermetrics has been acquired by webrtc.ventures
There is no team better equipped to continue the work of creating the most complete WebRTC monitoring tool out there.
Check out the project on github
github.com/peermetrics
There is no team better equipped to continue the work of creating the most complete WebRTC monitoring tool out there.
Check out the project on github
github.com/peermetrics
peer metrics
the complete WebRTC monitoring service. peer metrics has 5 repositories available. Follow their code on GitHub.
github.com
July 9, 2025 at 9:32 AM
Super happy to announce that peermetrics has been acquired by webrtc.ventures
There is no team better equipped to continue the work of creating the most complete WebRTC monitoring tool out there.
Check out the project on github
github.com/peermetrics
There is no team better equipped to continue the work of creating the most complete WebRTC monitoring tool out there.
Check out the project on github
github.com/peermetrics
@anthropic.com just launched a way to build and host AI-powered apps
If you're also wondering what AI-powered apps are, they're pretty much WebApps where the main focus is interacting or talking to an AI model.
www.anthropic.com/news/claude-...
If you're also wondering what AI-powered apps are, they're pretty much WebApps where the main focus is interacting or talking to an AI model.
www.anthropic.com/news/claude-...
Build and Host AI-Powered Apps with Claude - No Deployment Needed
Create interactive AI apps directly in Claude. Users pay for their own API usage, you pay nothing. Build games, tools, and assistants with zero deployment complexity.
www.anthropic.com
June 26, 2025 at 8:38 AM
@anthropic.com just launched a way to build and host AI-powered apps
If you're also wondering what AI-powered apps are, they're pretty much WebApps where the main focus is interacting or talking to an AI model.
www.anthropic.com/news/claude-...
If you're also wondering what AI-powered apps are, they're pretty much WebApps where the main focus is interacting or talking to an AI model.
www.anthropic.com/news/claude-...
Reposted by Andrei Onel
the level of misinformation sparked because of this bananas EEG preprint is just really tragic.
By the way if you think that a researcher caused cognitive decline to happen to participants in a study you should probably be freaked out by that
By the way if you think that a researcher caused cognitive decline to happen to participants in a study you should probably be freaked out by that
June 20, 2025 at 8:33 PM
the level of misinformation sparked because of this bananas EEG preprint is just really tragic.
By the way if you think that a researcher caused cognitive decline to happen to participants in a study you should probably be freaked out by that
By the way if you think that a researcher caused cognitive decline to happen to participants in a study you should probably be freaked out by that
Reposted by Andrei Onel
I regret nothing yet! fly.io/blog/youre-a...
My AI Skeptic Friends Are All Nuts
My smartest friends have bananas arguments about LLM coding.
fly.io
June 2, 2025 at 9:07 PM
I regret nothing yet! fly.io/blog/youre-a...
Really excited for new models from Anthropic. Can't wait to test them on creating documentation. They had an example in the demo.
www.youtube.com/live/EvtPBaa...
www.youtube.com/live/EvtPBaa...
Code with Claude Opening Keynote
YouTube video by Anthropic
www.youtube.com
May 22, 2025 at 5:48 PM
Really excited for new models from Anthropic. Can't wait to test them on creating documentation. They had an example in the demo.
www.youtube.com/live/EvtPBaa...
www.youtube.com/live/EvtPBaa...
Reposted by Andrei Onel
Product evals are misunderstood. Many teams think that adding another tool, metric, or llm-as-judge will solve all their problems and save their product. But that just dodges the hard truth and avoids the real work. Here's how to fix your process instead.
eugeneyan.com/writing/eval...
eugeneyan.com/writing/eval...
An LLM‑as‑Judge Won't Save Your Product—Fixing Your Process Will
Applying the scientific method, building via eval-driven development, and monitoring AI output.
eugeneyan.com
April 23, 2025 at 2:45 AM
Product evals are misunderstood. Many teams think that adding another tool, metric, or llm-as-judge will solve all their problems and save their product. But that just dodges the hard truth and avoids the real work. Here's how to fix your process instead.
eugeneyan.com/writing/eval...
eugeneyan.com/writing/eval...
𝗟𝗼𝗻𝗴 𝗰𝗼𝗻𝘁𝗲𝘅𝘁 𝘃𝘀 𝗥𝗔𝗚
With the launch of Llama 4 this weekend and its long context of 10 million, the old debate of "is RAG dead?" started to crop up again in the AI community.
With the launch of Llama 4 this weekend and its long context of 10 million, the old debate of "is RAG dead?" started to crop up again in the AI community.
April 10, 2025 at 9:49 AM
𝗟𝗼𝗻𝗴 𝗰𝗼𝗻𝘁𝗲𝘅𝘁 𝘃𝘀 𝗥𝗔𝗚
With the launch of Llama 4 this weekend and its long context of 10 million, the old debate of "is RAG dead?" started to crop up again in the AI community.
With the launch of Llama 4 this weekend and its long context of 10 million, the old debate of "is RAG dead?" started to crop up again in the AI community.
History Professor Answers Dictator Questions | Tech Support | WIRED
YouTube video by WIRED
youtu.be
April 7, 2025 at 1:21 PM
Reposted by Andrei Onel
Meta just dropped Llama 4 on a weekend! Two new open weight models (Scout and Maverick) and a preview of a model called Behemoth - Scout has a 10 million token context
Best information right now appears to be this blog post: ai.meta.com/blog/llama-4...
Best information right now appears to be this blog post: ai.meta.com/blog/llama-4...
The Llama 4 herd: The beginning of a new era of natively multimodal AI innovation
We’re introducing Llama 4 Scout and Llama 4 Maverick, the first open-weight natively multimodal models with unprecedented context support and our first built using a mixture-of-experts (MoE) architect...
ai.meta.com
April 5, 2025 at 7:53 PM
Meta just dropped Llama 4 on a weekend! Two new open weight models (Scout and Maverick) and a preview of a model called Behemoth - Scout has a 10 million token context
Best information right now appears to be this blog post: ai.meta.com/blog/llama-4...
Best information right now appears to be this blog post: ai.meta.com/blog/llama-4...
Reposted by Andrei Onel
The hardest part about finetuning is that people don't have labeled data. Today, @databricks.bsky.social introduced TAO, a new finetuning method that only needs inputs, no labels necessary. Best of all, it actually beats supervised finetuning on labeled data. www.databricks.com/blog/tao-usi...
TAO: Using test-time compute to train efficient LLMs without labeled data
LIFT fine-tunes LLMs without labels using reinforcement learning, boosting performance on enterprise tasks.
www.databricks.com
March 25, 2025 at 5:19 PM
The hardest part about finetuning is that people don't have labeled data. Today, @databricks.bsky.social introduced TAO, a new finetuning method that only needs inputs, no labels necessary. Best of all, it actually beats supervised finetuning on labeled data. www.databricks.com/blog/tao-usi...
Reposted by Andrei Onel
Search & recsys have historically drawn inspiration from language modeling, such as Word2vec for item embeddings, and GRUs and BERT for sequential models. LLMs are no different and have led to updates in model architecture, training data & paradigms, etc.
eugeneyan.com/writing/recs...
eugeneyan.com/writing/recs...
Improving Recommendation Systems & Search in the Age of LLMs
Model architectures, data generation, training paradigms, and unified frameworks inspired by LLMs.
eugeneyan.com
March 19, 2025 at 1:41 AM
Search & recsys have historically drawn inspiration from language modeling, such as Word2vec for item embeddings, and GRUs and BERT for sequential models. LLMs are no different and have led to updates in model architecture, training data & paradigms, etc.
eugeneyan.com/writing/recs...
eugeneyan.com/writing/recs...
Reposted by Andrei Onel
Two big LLM releases today - Cohere's Command A simonwillison.net/2025/Mar/13/... and Ai2's OLMo 2
OLMo claims to be "the first fully-open model (all data, code, weights, and details are freely available) to outperform GPT3.5-Turbo and GPT-4o mini", which feels notable allenai.org/blog/olmo2-32B
OLMo claims to be "the first fully-open model (all data, code, weights, and details are freely available) to outperform GPT3.5-Turbo and GPT-4o mini", which feels notable allenai.org/blog/olmo2-32B
Introducing Command A: Max performance, minimal compute
New LLM release from Cohere. It's interesting to see which aspects of the model they're highlighting, as an indicator of what their commercial customers value the most (highlight mine): > …
simonwillison.net
March 13, 2025 at 9:32 PM
Two big LLM releases today - Cohere's Command A simonwillison.net/2025/Mar/13/... and Ai2's OLMo 2
OLMo claims to be "the first fully-open model (all data, code, weights, and details are freely available) to outperform GPT3.5-Turbo and GPT-4o mini", which feels notable allenai.org/blog/olmo2-32B
OLMo claims to be "the first fully-open model (all data, code, weights, and details are freely available) to outperform GPT3.5-Turbo and GPT-4o mini", which feels notable allenai.org/blog/olmo2-32B
The pricing for the new OpenAI GPT 4.5 is crazy. They said it's a big model but holly.
We haven't seen this prices since 2021
We haven't seen this prices since 2021
February 28, 2025 at 9:00 AM
The pricing for the new OpenAI GPT 4.5 is crazy. They said it's a big model but holly.
We haven't seen this prices since 2021
We haven't seen this prices since 2021
Reposted by Andrei Onel
The new Anthropic Sonnet 3.7 is already available on askmanu.co.
The model tops the coding benchmarks and together with the most up to date docs from askmanu, you get he most powerful code assistant.
👇
The model tops the coding benchmarks and together with the most up to date docs from askmanu, you get he most powerful code assistant.
👇
askmanu
Code copilot that helps devs get answers fast
askmanu.co
February 25, 2025 at 9:42 AM
The new Anthropic Sonnet 3.7 is already available on askmanu.co.
The model tops the coding benchmarks and together with the most up to date docs from askmanu, you get he most powerful code assistant.
👇
The model tops the coding benchmarks and together with the most up to date docs from askmanu, you get he most powerful code assistant.
👇