In 30 seconds
Matthew Berman recounts a conversation with Thibault about increasing inference speeds and suggests that local computers may become a bottleneck, leading more agent work to move to the cloud. The short presents a forecast without benchmarks or a specified workload. Cloud inference and the environment in which an agent runs tools are separate choices, and either local or remote execution can be appropriate.
- 1An inference-speed observationBerman recalls Thibault discussing faster token generation and an anticipated ultra-fast mode; the clip supplies no measurements.
- 2A possible local bottleneckA local CPU could limit some tool-heavy workflows, but the short does not show that every ChatGPT user’s computer is too slow.
- 3A cloud forecastBerman expects more agent workloads to move to cloud environments, a prediction rather than a necessary consequence of faster inference.
Takeaways
- Faster model output can expose another bottleneck in an agent workflow, but the limiting component must be measured for that workload.
- Cloud model inference does not necessarily mean that the agent’s tools or code execute in the cloud.
- Cloud migration is Berman’s expectation, not a result established by the short.
Read the edited text by chapter
Faster inference and the local machine
Berman relates a conversation about rising token speeds and possible CPU bottlenecks. The short does not present a benchmark or identify a specific workload.
After a conversation with Thibault, I have been thinking about faster inference, measured in tokens per second, and an anticipated ultra-fast mode. The short does not provide a model, benchmark, release date or evidence for the title’s suggestion of infinite speed.
I suggest that the user’s computer and CPU could become the bottleneck as inference gets faster. That can happen for a particular agent workflow if local tool execution or processing is slow, but the clip provides no measurements to show that a typical ChatGPT client or all computers are CPU-bound. Network delays, server inference, tool calls, storage and other factors can also matter.
The case for cloud execution
Berman predicts more agent work will move to the cloud. The location of inference and the location of tool execution are separate decisions.
I expect more AI agent work to move to the cloud. That is a forecast, not a necessary consequence of higher token speeds. Cloud model inference and where an agent runs code or tools are different choices: [OpenAI’s architecture documentation](https://developers.openai.com/api/docs/guides/agents-api/architecture) describes execution environments that can be remote or local.
More from this channel
34:21EN
Matthew Berman· Oct 27, 2026Matthew Berman Examines Anthropic’s IPO Filing, AI Growth and GovernanceMatthew Berman, Sam Altman · Forward FutureThe presenter reviews the S-1’s financial metrics, compute strategy and voting structure while questioning the tension between rapid AI development and safety.
0:36ENMatthew Berman· Oct 25, 2026Kimi K3.1 and a Rubik’s Cube Solver DemonstrationMatthew Berman · Forward FutureMatthew Berman shows a scrambled virtual cube being solved and gives the Kimi K3.1 test a favorable assessment.
0:21ENMatthew Berman· Oct 24, 2026An AI-Generated Kimi Benchmark Chart Used Invented ScoresMatthew Berman · Forward FutureMatthew Berman asks for a graph of Kimi K3.1 Pro benchmarks, then discovers that the displayed figures are illustrative placeholders.
25:50ENMatthew Berman· Oct 24, 2026Kimi K3.1 Pro Tested: Strong 3D Demos, Uneven Results and Competitive PricingMatthew Berman · Forward FutureMatthew Berman tests the open-weight model on coding, 3D rendering and visual explanations, then compares benchmark results and prices.
32:20ENMatthew Berman· Oct 23, 2026Who Is Local AI For? The Costs, Trade-offs and a Hybrid FutureMatthew Berman · Forward FutureMatthew Berman weighs data control and independence against hardware cost, setup complexity and inference speed.
1:03:37ENMatthew Berman· Oct 23, 2026Matthew Berman’s Team Tests Kimi K3.1 on Animation, Games, and ChartsMatthew Berman, Alex, Brian · Forward FutureA live team AMA turns into an informal test of Kimi K3.1, from personal-agent workflows to motion graphics, a browser game, and a benchmark chart.