John Kim

@NoahZiems what are you looking forward to the most?

CAIS is going to be such a good conference...

@lateinteraction thanks omar!!!

@johnkimdw @NorthwesternU @ManlingLi_ wow congratulations John!!

I’m thrilled to share that I’ll be starting my CS PhD at @NorthwesternU this fall, advised by @ManlingLi_! I’ll be researching areas in trustworthy AI and spatial intelligence to build reliable AI systems that are grounded in the physical world. I’m also happy to announce that I was awarded the @NSF GRFP fellowship, which will support my PhD for 3 years! This wouldn’t have been possible without my wonderful mentors @nunompmoniz, @Meng_CS, @frank_liu_01, @NoahZiems, and countless others who’ve guided me throughout my undergrad. And so… I guess I won’t be leaving the midwest :)

@phuguo how did you use codex in the initial investigation?

Codex and I helped root cause goblins! We traced it to a reward signal intended to train the "Nerdy" personality - we found that it scored outputs with goblins higher, and as it boosted goblins in Nerdy training, the behavior generalized. See the blog post!

@wzenus congratulations!!

RAGEN-2 is selected as ICML 2026 spotlight (top 2.2%)!

@jaraxiong thank you!

@johnkimdw @NorthwesternU @ManlingLi_ Congrats!

@kexun_zhang @NorthwesternU @ManlingLi_ thank you kexun!!

@johnkimdw @NorthwesternU @ManlingLi_ Congratulations!

@ToadyBabirusa thank you for your concern, but i know shes an amazing advisor from meeting her, talking to her students, and other sources :)

@johnkimdw @NorthwesternU @ManlingLi_ Your advisor is recently on a black list for not treating students properly created by Chinese students. You may want to check that.

@sinankprn thanks!

@johnkimdw @NorthwesternU @ManlingLi_ All the best with it!

@NoahZiems Thank you Noah!!

Congrats John! Excited to see all the great things I'm sure you'll work on

@frank_liu_01 preprint: arxiv.org/pdf/2601.06757

excited to share our paper was accepted to #ACL2026 findings (w/ @frank_liu_01)! our benchmark evaluates the contextual safety of multimodal models across multi-turn conversations, showing models fail to balance safety awareness and helpfulness. looking forward to san diego :)

@gabriberton curious if you think youd stay in industry or return to academia?

A lot of people asked me about a postdoc at Amazon, so I jotted down my experience. Few companies have postdoc positions. Amazon, Meta and few others. These postdocs are purely with a company, unlike industrial PhDs which are affiliated with both a company and a university. Postdocs are very few, and very diverse. Amazon has ~30 postdocs. There are more VPs than postdocs! And Amazon is a huge company, some orgs are wildly different from others, so my experience sure is completely different from others'. Your org, the team, and the manager will make a huge difference. I had freedom on what research to work on, as long as it focused on VLMs and LLMs, and tackled challenges related to the product. I was in the Rufus team. I had to work on both product and research, but mostly research. I know other postdocs who did purely research. I had freedom to collaborate with other teams, and had a lot of chats / collabs with 10+ teams scattered across Amazon. That was great, I learned so much. The salary was good, like Applied Scientist kind of salary. I know most companies pay postdocs very little. Meta pays less than half than Amazon, for a postdoc in the Bay Area! One downside is that there is not much research going on at Amazon, most teams are focussed only on the product. Another downside (of working in any big tech) is that the bureaucracy is annoying. Every dataset needs to be approved by legal, which can require many steps, like double checking the source of the data, blurring all the faces, making sure there are no privacy issues. Publishing a paper must also be approved by legal (as in any big tech), which is something that people in academia don't need to worry about. Overall, if you have the chance I'd recommend a postdoc at Amazon. It's a short-term position (1 or 2 years) and it often it can be converted to a full time offer. It also opens doors both for academia and industry, while allowing you to live a comfortable life, which can't be done on an academic postdoc salary.

@akshatgupta57 @NoahZiems i understood it as in the future, there is going to be one “best” way to use coding agents, but rn we havent found that and are noisily converging to it

@NoahZiems I don’t understand 3. I feel like it could mean anything but I’m curious to hear what specific thing you were referring to

I'm convinced that: 1. No two people use coding agents in quite the same way 2. On the whole, everyone is moving in a similar direction 3. The combination of (1) and (2) suggests there is some place we all converge on that is wildly different from what we have right now.

@jacobli99 i just started using it and it makes so much more sense to have notes as graphs

obsidian is so cool

list of resources i found useful for post-training basics: - rlhfbook.com/book.pdf - huggingface.co/spaces/Hugging… - pytorch.org/blog/a-primer-… - thinkingmachines.ai/blog/lora/ - djdumpling.github.io/2026/01/31/fro… feel free to comment any others!

@ashertrockman dmed :)

I'm hiring a student researcher to work on RL and RLM-flavored things. DM me if interested

@loganthorneloe sent dm :)

My team at Google is hiring ML and software engineers in Pittsburgh and Mountain View! We're looking for someone who is: - Curious - High agency - Excited about solving difficult problems If this is you, send me a DM. (If you've already sent one and I didn't respond, ping me again!)

@ajeya_cotra dm sent :)

Come work with me! METR is looking for engineers, scientists, and a director of operations. Links to open posts in thread, and feel free to DM if you have questions!

@vincentweisser open to either MTS (full stack) or internship. > prev @ aws sagemaker building end-to-end multiagent systems for my service (data + ai dev env) > acl 2026 paper co-1st author submitted on multimodal contextual safety (arxiv.org/pdf/2601.06757) > tbd starting phd fall 2026 on building reliable ai systems (hence intern for this summer)

We are hiring for 19 open roles at @PrimeIntellect Join us to build the open superintelligence stack. - Member of Technical Staff - GPU Infrastructure - Member of Technical Staff - Full Stack - Member of Technical Staff - Inference - Account Executive - Compute Partnerships Lead - Founding GTM Lead - Head of Enterprise - Head of Growth - Revenue Operations - Business Operations Lead - Founder's Associate, Business Operations - Internship - Open Application for Unconventional Talent - Lead Design Engineer - AI Research Resident - Open Source AGI - Applied Research - Evals and Data - Applied Research - RL and Agents - Research Engineer - Distributed Training - Research Engineer - Reinforcement Learning Ping me or anyone on the team if you want to join us - share examples of exceptional work. $10K referral bonus. jobs.ashbyhq.com/PrimeIntellect