--- schema: 1 type: video title: 'Tokens Are the Cheap Part — Steve Jamison, Jellyfish' description: >- AI spending rises dramatically while the allocated cost of a merged pull request falls. Steve Jamison explains why those observations can coexist when developer salaries and model costs are counted together. Using Jellyfish's engineering data, he reports a 60 fold increase in AI spending, 43% more merged pull requests per engineer and a 26% decline in combined cost per merged pull request over the period discussed. The largest cost in that calculation remains salary. These are observational comparisons, rather than proof that token spending caused the throughput change or that companies realized equivalent cash savings. Jamison compares organizations with different levels of AI spending and then examines developers' weeks with and without heavier agent skill use. The analysis asks whether additional output spreads the salary allocation across enough work to outweigh the higher token bill. He also considers quality and work size: reverts offer one limited signal, while lines of code and pull request counts cannot by themselves establish customer value. A larger or more numerous set of changes may still need review, maintenance and evidence that it solves the right problem. The useful management question is therefore broader than whether the AI invoice went up. Teams need a consistent view of total engineering cost, output and quality before choosing where to encourage adoption or constrain spending. The talk provides a way to reason about that tradeoff while leaving causal attribution, economic value and the limits of output proxies open to scrutiny. Speaker info: - https://www.linkedin.com/in/stevetjamison/ Timestamps: 00:00 - AI spend and merged PR throughput move together 2:48 - Add developer salaries to the cost model 4:15 - Allocated cost per merged pull request 5:00 - Comparing lighter and heavier spending groups 8:55 - Comparing developers with their own prior weeks 10:12 - Reverts as a limited quality proxy 11:02 - Work size and lines of code 15:02 - Questions to ask before capping token spend canonical: >- https://liquidgold.cc/channel/aidotengineer/explore/video/tokens-are-the-cheap-part-steve-jamison-jellyfish-5f88c51f5935 markdown: >- https://liquidgold.cc/channel/aidotengineer/video/tokens-are-the-cheap-part-steve-jamison-jellyfish-5f88c51f5935.md llms: null parent: 'https://liquidgold.cc/channel/aidotengineer.md' channel: aidotengineer updated: '2026-10-23T22:24:08.864Z' language: null children: [] videoId: U7ZYBCcdJPg durationSeconds: 1052 publishedAt: '2026-10-13T17:00:19.000Z' searchChunks: - ordinal: 0 text: null startSeconds: 14 endSeconds: 171 watch: 'https://liquidgold.cc/watch/U7ZYBCcdJPg?t=14' heldReason: rights - ordinal: 1 text: null startSeconds: 14 endSeconds: 171 watch: 'https://liquidgold.cc/watch/U7ZYBCcdJPg?t=14' heldReason: rights - ordinal: 2 text: null startSeconds: 171 endSeconds: 183 watch: 'https://liquidgold.cc/watch/U7ZYBCcdJPg?t=171' heldReason: rights - ordinal: 3 text: null startSeconds: 183 endSeconds: 259 watch: 'https://liquidgold.cc/watch/U7ZYBCcdJPg?t=183' heldReason: rights - ordinal: 4 text: null startSeconds: 259 endSeconds: 270 watch: 'https://liquidgold.cc/watch/U7ZYBCcdJPg?t=259' heldReason: rights - ordinal: 5 text: null startSeconds: 270 endSeconds: 304 watch: 'https://liquidgold.cc/watch/U7ZYBCcdJPg?t=270' heldReason: rights - ordinal: 6 text: null startSeconds: 316 endSeconds: 535 watch: 'https://liquidgold.cc/watch/U7ZYBCcdJPg?t=316' heldReason: rights - ordinal: 7 text: null startSeconds: 316 endSeconds: 535 watch: 'https://liquidgold.cc/watch/U7ZYBCcdJPg?t=316' heldReason: rights - ordinal: 8 text: null startSeconds: 316 endSeconds: 535 watch: 'https://liquidgold.cc/watch/U7ZYBCcdJPg?t=316' heldReason: rights - ordinal: 9 text: null startSeconds: 547 endSeconds: 616 watch: 'https://liquidgold.cc/watch/U7ZYBCcdJPg?t=547' heldReason: rights - ordinal: 10 text: null startSeconds: 616 endSeconds: 629 watch: 'https://liquidgold.cc/watch/U7ZYBCcdJPg?t=616' heldReason: rights - ordinal: 11 text: null startSeconds: 662 endSeconds: 675 watch: 'https://liquidgold.cc/watch/U7ZYBCcdJPg?t=662' heldReason: rights - ordinal: 12 text: null startSeconds: 675 endSeconds: 902 watch: 'https://liquidgold.cc/watch/U7ZYBCcdJPg?t=675' heldReason: rights - ordinal: 13 text: null startSeconds: 912 endSeconds: 1052 watch: 'https://liquidgold.cc/watch/U7ZYBCcdJPg?t=912' heldReason: rights --- # Tokens Are the Cheap Part — Steve Jamison, Jellyfish AI spending rises dramatically while the allocated cost of a merged pull request falls. Steve Jamison explains why those observations can coexist when developer salaries and model costs are counted together. Using Jellyfish's engineering data, he reports a 60 fold increase in AI spending, 43% more merged pull requests per engineer and a 26% decline in combined cost per merged pull request over the period discussed. The largest cost in that calculation remains salary. These are observational comparisons, rather than proof that token spending caused the throughput change or that companies realized equivalent cash savings. Jamison compares organizations with different levels of AI spending and then examines developers' weeks with and without heavier agent skill use. The analysis asks whether additional output spreads the salary allocation across enough work to outweigh the higher token bill. He also considers quality and work size: reverts offer one limited signal, while lines of code and pull request counts cannot by themselves establish customer value. A larger or more numerous set of changes may still need review, maintenance and evidence that it solves the right problem. The useful management question is therefore broader than whether the AI invoice went up. Teams need a consistent view of total engineering cost, output and quality before choosing where to encourage adoption or constrain spending. The talk provides a way to reason about that tradeoff while leaving causal attribution, economic value and the limits of output proxies open to scrutiny. Speaker info: - https://www.linkedin.com/in/stevetjamison/ Timestamps: 00:00 - AI spend and merged PR throughput move together 2:48 - Add developer salaries to the cost model 4:15 - Allocated cost per merged pull request 5:00 - Comparing lighter and heavier spending groups 8:55 - Comparing developers with their own prior weeks 10:12 - Reverts as a limited quality proxy 11:02 - Work size and lines of code 15:02 - Questions to ask before capping token spend Canonical: https://liquidgold.cc/channel/aidotengineer/explore/video/tokens-are-the-cheap-part-steve-jamison-jellyfish-5f88c51f5935 · Channel: [AI Engineer](https://liquidgold.cc/channel/aidotengineer.md) > Search chunk text is held until public-source provenance is verified. ## Document chunks --- #### Chunk 0: Document excerpt (0:02 to 2:51) > Held: transcript rights do not permit this chunk's text. --- #### Chunk 1: Document excerpt (2:51 to 4:19) > Held: transcript rights do not permit this chunk's text. --- #### Chunk 2: Document excerpt (4:19 to 5:04) > Held: transcript rights do not permit this chunk's text. --- #### Chunk 3: Document excerpt (5:04 to 8:55) > Held: transcript rights do not permit this chunk's text. --- #### Chunk 4: Document excerpt (8:55 to 10:16) > Held: transcript rights do not permit this chunk's text. --- #### Chunk 5: Document excerpt (10:16 to 11:02) > Held: transcript rights do not permit this chunk's text. --- #### Chunk 6: Document excerpt (11:02 to 15:02) > Held: transcript rights do not permit this chunk's text. --- #### Chunk 7: Document excerpt (15:02 to 17:32) > Held: transcript rights do not permit this chunk's text.