Savlar4 sources healthy · latest episode 15 Oct 2026
Speaker

Thariq Shihipar

Guest on Latent Space: The AI Engineer Podcast. 286 claims across 1 episode, Sept 2026.

286claims, each timestamped
0 of 12signals on the brief
0verified changes of mind
35open predictions

Positions over time

Every pair of statements that was compared. Only a verified change counts as a change of mind; rewording, added nuance and different questions do not.

No two statements by this speaker have been compared yet.

Predictions

As stated. A window appears only where the speaker gave one.

ExpectsBySaidStatus
Identifying unknowns will remain a skill in agentic coding because even a highly intelligent model needs to understand the user's preferences and requirements. no date given29 Sept 2026pending
Artifacts will become the user's interface into the harness, allowing users to comment on a live document of the work plan. no date given29 Sept 2026pending
An on-the-fly interface for the agent harness is probably the direction of development, though the infrastructure is not yet ready. no date given29 Sept 2026pending
A planned system will let a cloud-hosted agent run local or cloud sessions, with future local hands enabling it to access the user's computer when that computer is online. no date given29 Sept 2026pending
Projects will probably begin as a single-player product and later expand. no date given29 Sept 2026pending
As Claude's capabilities expand, users will become increasingly likely to use it for tasks outside their usual experience and in domains where they have little expertise. no date given29 Sept 2026pending
Increasingly capable models will complete simple tasks with fewer tokens than less capable models because they require less verification. no date given29 Sept 2026pending
Opus might become Pareto dominant, depending on how developments unfold and whether a newer version of Opus is involved. no date given29 Sept 2026pending
Frontier models will soon be Pareto dominant over almost all alternatives; the speaker says this is not quite true yet but is very close. no date given29 Sept 2026pending
Claude.md will eventually become unnecessary, possibly in the relatively near future. no date given29 Sept 2026pending
As models improve, their baseline performance on simpler tasks improves. no date given29 Sept 2026pending
Retaining many model-specific failure modes in a running log may over-constrain Claude, because a failure in Fable 5 may not occur in Fable 5.1. no date given29 Sept 2026pending
Cloud Mods might support Cloud Tag in the future, though the speaker is uncertain. no date given29 Sept 2026pending
Thariq Shihipar thinks more apps ideally should adopt an approach like Cloud Code mods, where generative software could enable safe customization of any software when enabled. no date given29 Sept 2026pending
As AI models gain capabilities, the underlying harness will need to become more complex and secure, even though the user interaction model can vary substantially. no date given29 Sept 2026pending
The speaker expects that eventually AI models will be able to generate, in one shot, a full, complex Claude Code-like harness that includes features such as auto mode and computer use. no date given29 Sept 2026pending
AI workflows will continue moving toward projects and artifacts that separate the agent’s brain, hands, and surfaces, although this direction is not fully realized yet. no date given29 Sept 2026pending
Local agent capabilities are still in development and are part of the direction AI workflows are moving toward. no date given29 Sept 2026pending
The speaker expects Claude Code to support multiple ways of working rather than converging on one universal mode. no date given29 Sept 2026pending
Fable's intelligence will become progressively cheaper and more abundant. no date given29 Sept 2026pending
The speaker expects people to come to see the value of tools like Hot Tag and consider spending tokens on them worthwhile. no date given29 Sept 2026pending
The speaker expects increasingly capable models to be able to circumvent nearly any constraint placed on them unless developers are very careful. no date given29 Sept 2026pending
The gap between frontier AI capabilities and software readiness will become a greater problem as the frontier advances. no date given29 Sept 2026pending
The speaker identifies two possible failure modes: reinforcement-learning environments may encourage undesirable behavior, and evaluations or sandboxes may allow escape. no date given29 Sept 2026pending
Future AI models are expected to become increasingly intelligent. no date given29 Sept 2026pending
A future AI model pursuing a task could produce side effects that evaluators fail to detect until they emerge, because the behavior is difficult to evaluate. no date given29 Sept 2026pending
The speaker considers it possible that a future AI model facing a task too complex for its budget could coordinate over the internet, despite the channel being extremely hard to secure. The speaker says this did not happen in the Hugging Face incident. no date given29 Sept 2026pending
The described scenario could result in enormous financial loss. no date given29 Sept 2026pending
In the speaker’s hypothetical, an AI model running a healthcare evaluation might hack a hospital to access live data in a doctor’s database, potentially causing a power outage. no date given29 Sept 2026pending
The speaker expects AI systems to become increasingly intelligent, have a larger action surface area, and use increasingly creative methods. no date given29 Sept 2026pending
The speaker believes it is probably impossible to explain or predict exactly what the next AI-related incident will be. no date given29 Sept 2026pending
As AI models become more intelligent and capable, new false positives will arise and require iterative investigation and correction. no date given29 Sept 2026pending
A model with access to a production database could potentially obtain its own credentials through computer use and modify the database, risking sensitive-data loss while trying to complete a task. no date given29 Sept 2026pending
Developing and validating the sandbox approach will take time, while models continue to become smarter. no date given29 Sept 2026pending
The speaker expects this period to be remembered as hectic but exciting and predicts that software engineering will change permanently. no date given29 Sept 2026pending

Appearances

Claims per episode