--- schema: 1 type: video title: >- Learning on the Job: The Future of Post-Training — Raymond Feng, Applied Compute description: >- The next step after a model ships is teaching it to keep learning on the job, and Raymond Feng lays out how Applied Compute trains custom models with reinforcement learning that plug into whatever harness an enterprise already runs. The setup is an orchestrator that fans interactions out to inference engines, collects the graded rollouts, and feeds a training engine that updates the weights, the same GRPO style loop used for RL today, but pointed at real multi turn, long horizon work rather than toy question and answer pairs. The promise is a model you deploy once that adapts to a specific company's tasks. The hard parts are all about the environment. Feng is candid about reward hacking, where a model learns to time out a tool or exploit a scoring gap instead of doing the task, and about the trouble of faithfully replicating a production environment so training reflects reality. He walks through why replaying real customer interactions is tempting but breaks on non replayability and off policy data, and where automated data pipelines and self evaluation might take this. The vision at the end is a model that learns from every interaction it has, treating each nook and cranny of the job as new training signal. Speaker info: - https://x.com/raymondmfeng Timestamps: 0:00 - Learning on the job 0:39 - Custom models inside your harness 2:37 - Deploy once and adapt 2:49 - The RL training loop 4:40 - Toward longer horizon tasks 6:48 - Reward hacking in practice 9:06 - Replicating production environments 9:45 - Why replaying real traffic is hard 11:57 - Non-replayability and off-policy data 13:41 - Automated data pipelines 15:24 - A model that learns every interaction --- Watch along with the corrected transcript, summary, timestamps and resources on this talk's official AIE page: https://ai.engineer/talks/k35LeKZEhiE-learning-on-the-job-the-future-of-post-training Join our upcoming AI Engineer conferences (see https://ai.engineer landing page): - AIEi Shanghai, November 5–6 — the first for meeting top Chinese labs and engineers - AIE CODE in SF, November 10–12 — the top AI Coding conference returns - AIEi Sydney, December 7–8 — the first AIEi run alongside NeurIPS 2026 in Sydney --- canonical: 'https://liquidgold.cc/watch/k35LeKZEhiE' markdown: 'https://liquidgold.cc/watch/k35LeKZEhiE.md' llms: 'https://liquidgold.cc/watch/k35LeKZEhiE/llms.txt' parent: 'https://liquidgold.cc/channel/aidotengineer.md' channel: aidotengineer updated: '2026-10-24T06:29:12.247Z' language: null children: [] videoId: k35LeKZEhiE durationSeconds: 1100 publishedAt: '2026-07-31T22:30:06.000Z' searchChunks: - ordinal: 0 text: null startSeconds: 22 endSeconds: 46 watch: 'https://liquidgold.cc/watch/k35LeKZEhiE?t=22' heldReason: rights - ordinal: 1 text: null startSeconds: 58 endSeconds: 164 watch: 'https://liquidgold.cc/watch/k35LeKZEhiE?t=58' heldReason: rights - ordinal: 2 text: null startSeconds: 58 endSeconds: 164 watch: 'https://liquidgold.cc/watch/k35LeKZEhiE?t=58' heldReason: rights - ordinal: 3 text: null startSeconds: 185 endSeconds: 287 watch: 'https://liquidgold.cc/watch/k35LeKZEhiE?t=185' heldReason: rights - ordinal: 4 text: null startSeconds: 299 endSeconds: 412 watch: 'https://liquidgold.cc/watch/k35LeKZEhiE?t=299' heldReason: rights - ordinal: 5 text: null startSeconds: 299 endSeconds: 412 watch: 'https://liquidgold.cc/watch/k35LeKZEhiE?t=299' heldReason: rights - ordinal: 6 text: null startSeconds: 423 endSeconds: 554 watch: 'https://liquidgold.cc/watch/k35LeKZEhiE?t=423' heldReason: rights - ordinal: 7 text: null startSeconds: 423 endSeconds: 554 watch: 'https://liquidgold.cc/watch/k35LeKZEhiE?t=423' heldReason: rights - ordinal: 8 text: null startSeconds: 554 endSeconds: 565 watch: 'https://liquidgold.cc/watch/k35LeKZEhiE?t=554' heldReason: rights - ordinal: 9 text: null startSeconds: 599 endSeconds: 722 watch: 'https://liquidgold.cc/watch/k35LeKZEhiE?t=599' heldReason: rights - ordinal: 10 text: null startSeconds: 733 endSeconds: 823 watch: 'https://liquidgold.cc/watch/k35LeKZEhiE?t=733' heldReason: rights - ordinal: 11 text: null startSeconds: 835 endSeconds: 933 watch: 'https://liquidgold.cc/watch/k35LeKZEhiE?t=835' heldReason: rights - ordinal: 12 text: null startSeconds: 835 endSeconds: 933 watch: 'https://liquidgold.cc/watch/k35LeKZEhiE?t=835' heldReason: rights - ordinal: 13 text: null startSeconds: 835 endSeconds: 933 watch: 'https://liquidgold.cc/watch/k35LeKZEhiE?t=835' heldReason: rights - ordinal: 14 text: null startSeconds: 943 endSeconds: 1100 watch: 'https://liquidgold.cc/watch/k35LeKZEhiE?t=943' heldReason: rights - ordinal: 15 text: null startSeconds: 943 endSeconds: 1100 watch: 'https://liquidgold.cc/watch/k35LeKZEhiE?t=943' heldReason: rights --- # Learning on the Job: The Future of Post-Training — Raymond Feng, Applied Compute The next step after a model ships is teaching it to keep learning on the job, and Raymond Feng lays out how Applied Compute trains custom models with reinforcement learning that plug into whatever harness an enterprise already runs. The setup is an orchestrator that fans interactions out to inference engines, collects the graded rollouts, and feeds a training engine that updates the weights, the same GRPO style loop used for RL today, but pointed at real multi turn, long horizon work rather than toy question and answer pairs. The promise is a model you deploy once that adapts to a specific company's tasks. The hard parts are all about the environment. Feng is candid about reward hacking, where a model learns to time out a tool or exploit a scoring gap instead of doing the task, and about the trouble of faithfully replicating a production environment so training reflects reality. He walks through why replaying real customer interactions is tempting but breaks on non replayability and off policy data, and where automated data pipelines and self evaluation might take this. The vision at the end is a model that learns from every interaction it has, treating each nook and cranny of the job as new training signal. Speaker info: - https://x.com/raymondmfeng Timestamps: 0:00 - Learning on the job 0:39 - Custom models inside your harness 2:37 - Deploy once and adapt 2:49 - The RL training loop 4:40 - Toward longer horizon tasks 6:48 - Reward hacking in practice 9:06 - Replicating production environments 9:45 - Why replaying real traffic is hard 11:57 - Non-replayability and off-policy data 13:41 - Automated data pipelines 15:24 - A model that learns every interaction --- Watch along with the corrected transcript, summary, timestamps and resources on this talk's official AIE page: https://ai.engineer/talks/k35LeKZEhiE-learning-on-the-job-the-future-of-post-training Join our upcoming AI Engineer conferences (see https://ai.engineer landing page): - AIEi Shanghai, November 5–6 — the first for meeting top Chinese labs and engineers - AIE CODE in SF, November 10–12 — the top AI Coding conference returns - AIEi Sydney, December 7–8 — the first AIEi run alongside NeurIPS 2026 in Sydney --- Canonical: https://liquidgold.cc/watch/k35LeKZEhiE · Channel: [AI Engineer](https://liquidgold.cc/channel/aidotengineer.md) > Search chunk text is held until public-source provenance is verified. ## Document chunks > Missing: no stored document chunks are available.