30초 요약
Theo distinguishes the earlier Mythos preview from Mythos 5 and argues that Fable 5 uses the same underlying model with additional checks. He illustrates his interpretation with an analogy of two doors offering different access to the same place.
- 1From preview to releaseTheo describes Mythos preview as the model used during Project Glass Wing and says subsequent reinforcement learning led to Mythos 5.
- 2Two doorsTheo argues that Fable 5 and Mythos 5 are the same model reached through different access paths.
- 3Different checksHis analogy compares authorized direct access to a door with additional guards checking requests.
핵심 정리
- Theo distinguishes the Mythos preview used in Project Glass Wing from the later Mythos 5 release.
- Theo describes Fable 5 and Mythos 5 as two access paths to the same underlying model.
- His analogy contrasts direct access for authorized Mythos users with additional checks for Fable.
챕터별로 다듬은 전문 읽기
Preview and release
Theo distinguishes Mythos preview from Mythos 5. He describes the latter as the later production model.
Let's distinguish Fable 5 and Mythos 5. Mythos preview was used during Project Glass Wing; I understand Mythos 5 as the subsequent model produced through reinforcement learning, now presented as ready for production.
Two access paths
He argues that Fable 5 uses the same model with additional checks. He uses a two-door analogy to explain his view.
My view is that Fable 5 is the same underlying model, with a different way in. Imagine two doors to the same place: the Mythos 5 door lets someone with the right key enter directly, while the Fable 5 door has guards checking what the person is doing. I am concerned about how those checks work.
같은 채널의 다른 글
38:47EN
Theo - t3․gg· 2026년 10월 28일Theo’s Long-Running Agent Experiments: Verification, Compute and CostTheo Browne · Ping LabsA skateboard game and two compiler projects show how repeated checks and flexible compute can sustain agent work—and how expensive those loops can become.
36:07ENTheo - t3․gg· 2026년 10월 27일Theo on Token Costs, Coding Agents, and the Price of AI CapabilityTheo Browne · Ping LabsA viral estimate of Lauren's AI spending prompts a closer look at cache pricing, benchmarks, and the changing work of coding agents.
30:37ENTheo - t3․gg· 2026년 10월 26일Theo’s Three Skills for Reviewing and Shipping with Coding AgentsTheo Browne · Ping LabsHow sol-loop, full-send and full-stop encode review, verification and different levels of merge authority.
0:43ENTheo - t3․gg· 2026년 10월 23일Theo Distinguishes Security Vulnerabilities from Active AttacksTheo Browne · Ping LabsA short explanation of why disclosed software flaws and reported attacks are different kinds of security news.
29:22ENTheo - t3․gg· 2026년 10월 22일Theo on Security Vulnerabilities, Stolen API Keys and AITheo Browne · Ping LabsA look at the gap between vulnerability disclosure and patching, and a speculative account of why stolen inference keys may be valuable.
31:13ENTheo - t3․gg· 2026년 10월 20일Theo on Why Smarter Models Help More Than DevelopersTheo Browne · Ping LabsFrom game modifications to compiler experiments, Theo revisits his assumptions about who benefits from more capable AI.