cd/sources/vincentschmalbach-auto-discovered· home sources Vincentschmalbach (auto-discovered)
cat /sources/vincentschmalbach-auto-discovered.feed | wc -l → 47

Vincentschmalbach (auto-discovered)

articles 47 domain vincentschmalbach.com → page 1/3 feed RSS
11:25
2026-08-11
vincentschmalbach.com
large-language-models

How Floating-Point Determinism Affects LLM Reproducibility

Floating-point arithmetic non-determinism can cause LLM inference to produce different outputs across runs, hardware, or provider routing, according to an analysis of numerical execution in large lang…

11:25
2026-08-11
vincentschmalbach.com
large-language-models

Can Provider Routing Change LLM Outputs?

Provider routing can change an LLM's output when a request reaches a different model version, fallback model, parameter configuration, precision level, inference engine, region, or runtime environment…

11:25
2026-08-11
vincentschmalbach.com
large-language-models

How Model Updates Break LLM Reproducibility

A hosted LLM's output can change when providers update models, safety controls, routing, or serving infrastructure without code changes, breaking reproducibility. Google's Gemini API documentation exp…

11:25
2026-08-11
vincentschmalbach.com
large-language-models

What Is a System Fingerprint in LLM APIs?

OpenAI's API documentation explains that the `system_fingerprint` field in LLM responses is a provider-generated marker for the backend configuration serving a request, not a reflection of the user's …

11:25
2026-08-11
vincentschmalbach.com
large-language-models

Can Tokenizer Changes Affect LLM Output?

Changing a tokenizer can alter an LLM's output, ranging from no visible difference to lower quality, different formatting, shorter usable context, or complete inference failure, according to Hugging F…

11:25
2026-08-11
vincentschmalbach.com
large-language-models

Is the Same Prompt Always the Same LLM Input?

The same visible prompt is not always the same input received by a large language model (LLM), according to an analysis of provider routing and request assembly. Identical visible prompts do not estab…

11:25
2026-08-11
vincentschmalbach.com
artificial-intelligence

How to Test a Nondeterministic LLM Application

Testing a nondeterministic LLM application requires treating it as a workflow with behavioral requirements across repeated runs, not as a function returning one exact string, according to a guide that…

09:13
2026-08-11
vincentschmalbach.com
large-language-models

What Is Batch Invariance in LLM Inference?

VLLM's batch-invariance documentation defines a feature ensuring a request produces the same inference result regardless of batch size, composition, request order, or scheduling under a fixed hardware…

09:13
2026-08-11
vincentschmalbach.com
large-language-models

Do Structured Outputs Make LLM Responses Deterministic?

OpenAI reported that its strict Structured Outputs feature achieved 100% schema-matching reliability in an internal evaluation for gpt-4o-2024-08-06, but the company and other providers such as Amazon…

09:13
2026-08-11
vincentschmalbach.com
large-language-models

How Does Mixture-of-Experts Routing Affect LLM Repeatability?

A new analysis from the Journal of Machine Learning Research explains that Mixture-of-Experts (MoE) routing can cause large language models to produce different outputs across runs due to discrete exp…

09:12
2026-08-11
vincentschmalbach.com
large-language-models

Does Setting a Seed Make LLM Output Reproducible?

Setting a seed does not guarantee identical large language model (LLM) output, according to an analysis of LLM inference. A seed initializes the pseudorandom number generator used in token sampling bu…

05:34
2026-08-11
vincentschmalbach.com
large-language-models

How to Roll Out a New LLM Model Version Safely

A new large language model (LLM) version can alter production behavior in unpredictable ways, so organizations should use a gated, progressive, reversible rollout that includes defining a production c…

08:23
2026-08-10
vincentschmalbach.com
artificial-intelligence

Switching from GPT-5.5 to GPT-5.6 Made Me Less Productive

A developer reports that switching from OpenAI's GPT-5.5 to GPT-5.6 in Codex, the company's AI coding agent, has made him less productive, with tasks spinning indefinitely and draining his three $200-…

13:22
2026-08-06
vincentschmalbach.com
artificial-intelligence

AI Is Now a Commodity

AI has become a commodity, with DeepSeek V4 Flash costing $0.13 per million input tokens and $0.26 per million output tokens via OpenRouter, and TIME is serving AI crawlers a stripped-down markdown ve…

12:41
2026-08-05
vincentschmalbach.com
ai-policy

TIME Is Serving AI Bots a Different Website, With Ads Built In

TIME is serving AI crawlers a stripped-down markdown version of its website with ads embedded, while human visitors receive the full HTML page, according to a test by an unnamed researcher. The markdo…

13:21
2026-08-03
vincentschmalbach.com
large-language-models

GPT-5.6 Sol Uses Twice the Tokens of GPT-5.5

OpenAI's GPT-5.6 Sol xhigh uses 16.45M tokens per session, 2.25 times the 7.30M tokens per session of GPT-5.5 xhigh, according to a developer's analysis of 1,715 sessions over 14 days. Total token usa…

12:55
2026-08-03
vincentschmalbach.com
developer-tools

Codex CLI Has Responsive Tables in the Terminal

OpenAI's Codex CLI now renders responsive Markdown tables in the terminal, adapting to window width by switching from a grid to a record layout when columns are too narrow. The first implementation la…

page 1 / 3 next →