Résultats
31 articlesTrier
0:22EN
Matthew Berman· 18 oct. 2026DeepSeek's Memory Use: A Short Claim About KV CacheMatthew Berman · Forward FutureMatthew Berman links lower cache memory use to serving more tokens, but the excerpt does not state the comparison behind its figures.
0:50ENMatthew Berman· 2 sept. 2026Matthew Berman on Using Idle Computers for AI InferenceMatthew Berman · Forward FutureBerman points to Darkbloom as a way to put spare computing capacity to work on AI models, while raising the question of how far distributed inference can go.
0:41ENMatthew Berman· 28 août 2026Could Faster AI Agents Shift More Work to the Cloud?Matthew Berman · Forward FutureMatthew Berman reflects on a conversation about faster inference and local-computer limits. Whether the CPU is a bottleneck depends on the particular agent workload and execution environment.
0:48ENMatthew Berman· 25 août 2026Matthew Berman on Earning From Idle Mac Compute With DarkbloomMatthew Berman · Forward FutureA proposed network of Apple Silicon Macs could serve AI inference, but the earnings shown are projections that depend on hardware, availability and demand.
1:24ENMatthew Berman· 22 juil. 2026Matthew Berman on Using Model Routing to Reduce AI CostsMatthew Berman · Forward FutureA short, sponsored explanation of splitting planning and coding between models of different cost, with a customer savings claim from DigitalOcean.
0:24ENMatthew Berman· 2 mars 2026A Surprise in AI Token Usage and CostsMatthew Berman · Forward FutureMatthew Berman reacts to a rising usage bill and notes why monitoring token consumption matters.
0:52ENMatthew Berman· 13 févr. 2026GPT-5.3-Codex-Spark Brings Faster Coding InferenceMatthew Berman · Forward FutureMatthew Berman highlights OpenAI's smaller coding model and its use of Cerebras hardware for low-latency responses.
0:24ENMatthew Berman· 26 janv. 2026Matthew Berman Checks His Clawdbot Token CostsMatthew Berman · Forward FutureA short usage check reveals unexpectedly high model spending over two days.
12:13ENMatthew Berman· 15 janv. 2026OpenAI’s Cerebras Deal and the Race for Faster AI InferenceMatthew Berman, Andrew Feldman, Greg Brockman · Forward FutureMatthew Berman explains why inference speed and computing capacity matter, while a closer look at the agreement and benchmark puts his 100-times-faster scenario in context.
1:38ENMatthew Berman· 29 déc. 2025NVIDIA’s Groq Licensing Deal and the Inference MarketMatthew Berman · Forward FutureMatthew Berman discusses why NVIDIA licensed Groq’s inference technology and hired members of its team.
1:02ENMatthew Berman· 31 oct. 2025Inside a Cerebras Data Center in Oklahoma CityMatthew Berman · Forward FutureMatthew Berman tours the facility and explains its wafer-scale chips, racks and cooling system.
8:47ENMatthew Berman· 16 sept. 2025Why Identical LLM Prompts Can Produce Different Answers at Temperature ZeroMatthew Berman · Forward FutureMatthew Berman explains Thinking Machines Lab’s work on batch-invariant inference and what its reproducibility experiment actually showed.
12:10ENMatthew Berman· 6 mars 2025Mercury Coder and Diffusion Language Models: Speed, Demos and LimitsMatthew Berman · Forward FutureMatthew Berman explores Inception’s 2025 coding model, comparing iterative text refinement with autoregressive generation and examining speed claims.
10:01ENMatthew Berman· 5 févr. 2025DeepSeek R1 Code Optimization and a $3 Visual Counting ExperimentMatthew Berman · Forward FutureMatthew Berman examines a human-guided coding improvement and a small reinforcement-learning experiment, then considers their implications for AI research.
30:29ENMatthew Berman· 3 janv. 2025How Researchers Think o1-Style Reasoning Models WorkMatthew Berman · Forward FutureMatthew Berman explains a research survey of policy initialization, rewards, search and reinforcement learning, and the open questions behind reasoning models.
10:39ENMatthew Berman· 29 oct. 2024Zamba2-7B: A Hybrid Mamba–Transformer Model Tested on Coding and ReasoningMatthew Berman · Forward FutureMatthew Berman compares Zyphra’s efficiency claims with his own informal tests, including Python games, spatial puzzles and simple word tasks.
9:38ENMatthew Berman· 15 oct. 2024Liquid Foundation Models: Company Benchmarks and a Hands-On TestMatthew Berman · Forward FutureMatthew Berman reviews Liquid AI’s 2024 model and memory claims, then tries the 40B preview on coding, logic, and instruction-following prompts.
0:54ENMatthew Berman· 16 sept. 2024OpenAI’s o1 Preview: A First Look at Longer ReasoningMatthew Berman · Forward FutureMatthew Berman introduces the o1-preview and o1-mini models and tries a Python game prompt.
18:37ENMatthew Berman· 29 août 2024GameNGen Simulates Doom as AI Coding, Safety and Inference News AccelerateMatthew Berman · Forward FutureMatthew Berman reviews a 2024 game-simulation paper, Cursor, robotics, AI safety debates and new enterprise tools.
0:44ENMatthew Berman· 29 août 2024NVIDIA Adds Multi-LoRA Serving to Its RTX AI ToolkitMatthew Berman · Forward FutureIn an NVIDIA-partnered video, Matthew Berman describes using multiple adapters with one base model and a vendor-reported throughput gain.
5:50ENMatthew Berman· 27 juil. 2024A Mixture-of-Agents Demo Using Groq and StreamlitMatthew Berman · Forward FutureMatthew Berman walks through a 2024 Groq-powered multi-agent project, from setup to inspecting its layered responses.
11:20ENMatthew Berman· 1 juil. 2024Matthew Berman Adapts Mixture of Agents to Groq InferenceMatthew Berman · Forward FutureA 2024 tutorial changes an open-source multi-model ensemble to use Groq, then tests its interactive output and fixes a streaming error.
13:18ENMatthew Berman· 20 juin 2024Testing Local AI on a Dual RTX 6000 WorkstationMatthew Berman · Forward FutureMatthew Berman tests language models, concurrent inference, and image generation on a Dell Precision tower supplied by Dell and NVIDIA.
12:51ENMatthew Berman· 21 avr. 2024Llama 3 70B on Groq: Fast Inference and an Informal Prompt TestMatthew Berman · Forward FutureMatthew Berman tests coding, reasoning and repeated prompts on Groq’s April 2024 Llama 3 service, with mixed results and several scoring caveats.Charger plus24 sur 31 affichés