4 sources healthy · latest episode 9 Nov 2026
Speaker

Alessio Fanelli

Host of Latent Space: The AI Engineer Podcast; guest on Latent Space: The AI Engineer Podcast. 43 claims across 9 episodes, Apr 2025 to Sep 2026.

43claims, each timestamped
0 of 12signals on the brief
0verified changes of mind, of 5 pairs compared
3open 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.

  • NOT COMPARABLE · 0.9827 Apr → 29 SepThen GPT-4 and possibly Mini used defensive checks before acting, but the checks were often incorrect and sometimes hindered the models.Now Under programs such as Glasswing, security experts receive early, time-limited model access to conduct their own red-teaming before broader release.The earlier claim concerns defensive checks in GPT-4 and possibly Mini, including that they were sometimes incorrectly configured and hindered the models. The later claim describes Glasswing’s time-limited early access for security experts to red-team models. These are different mechanisms and claims; the later statement does not say whether the speaker’s view of the earlier defensive checks has changed.
  • NOT COMPARABLE · 0.9827 Apr → 29 SepThen Lab play evaluates models' spatial reasoning, API use, and performance in constrained environments.Now Under programs such as Glasswing, security experts receive early, time-limited model access to conduct their own red-teaming before broader release.The earlier claim says Lab Play measures models’ “spatial reasoning, API use, and performance in constrained environments.” The later claim describes giving security experts “early, time-limited model access” to red-team models. These concern different activities, so the later statement does not establish a changed position on the earlier claim.
  • NOT COMPARABLE · 0.9927 Apr → 25 JunThen The speaker says Factorio's tutorial made the game's complexity and potential as a learning environment apparent.Now The speaker considers Warp Code's AI intelligence potentially more valuable than Warp's terminal.These claims address different subjects: the earlier claim says the Factorio tutorial revealed the game's complexity and potential as a learning environment; the later claim says Warp Code's AI intelligence may be more valuable than Warp's terminal. They do not establish a change in position on the same question.
  • NOT COMPARABLE · 0.9827 Apr → 25 JunThen The speaker does not think AI agents have an inherent mechanical advantage over humans in Factorio.Now The speaker considers Warp Code's AI intelligence potentially more valuable than Warp's terminal.The earlier claim concerns whether AI agents have an inherent mechanical advantage over humans in Factorio. The later claim concerns whether Warp Code’s AI intelligence may be more valuable than Warp’s terminal. These address different questions, so the later statement does not establish a change in the earlier position.
  • NOT COMPARABLE · 0.9927 Apr → 25 JunThen The speaker thinks the largest potential improvement in Factorio agents depends on the structures they can devise.Now The speaker considers Warp Code's AI intelligence potentially more valuable than Warp's terminal.These claims concern different questions. The earlier claim is about Factorio agents and says improvement depends on “what type of structure are they able to come up with”; the later claim compares the value of Warp Code’s AI intelligence with Warp’s terminal. They do not establish a change in the speaker’s position on the same issue.

Predictions, in their words

As stated, each with its moment. A date appears only where the speaker gave one.

“Now, of course, as we go on, the models get better and better. It might improve in the future, but that's at least the current state that we found.”

Expects: Vision might improve results as models improve in the future. · no date given · not yet resolved

Alessio FanelliLatent Space: The AI Engineer Podcast · 27 Apr 2025
15:29Report a misattribution
“objective then probably we will we might see some wacky structures which maybe humans haven't thought of in the future but that's that's the the the gap towards there and the road towards there is long also guys well this was great and there's a lot of factorial fans as we wrap up a couple things are there any games that you wish people would build similar experiences on just to see how the models perform maybe can be you know”

Expects: Factorio agents trained toward the relevant objective might eventually devise unusual structures that humans have not thought of, but reaching that point will take a long time. · no date given · not yet resolved

Alessio FanelliLatent Space: The AI Engineer Podcast · 27 Apr 2025
25:16Report a misattribution
“Hopefully you can expand programs like that help.”

Expects: The speaker hopes that programs providing early access to models for security testing can be expanded to help address safety problems. · no date given · not yet resolved

Alessio FanelliLatent Space: The AI Engineer Podcast · 29 Sep 2026
1:27:00Report a misattribution

Appearances

Claims per episode