AI Tips Today · Monday, Wednesday, Friday · Nothing disappears
Practical, tested tips for getting more out of the frontier models you actually use — Claude, GPT, Gemini and whatever ships next. Written by the same newsroom, held to the same no hype standard, and archived permanently.
Wednesday, September 16, 2026 · Gemini 3.8 Live Extended Thinking (Google)
Google's new reasoning voice model refuses synchronous tool calls outright and replaces the turn complete signal with a session status field. Two small breaking changes that will quietly kill a working voice agent if you swap the model ID and nothing else.
Read the tip →Monday, September 14, 2026 · Gemini 3.8 Flash (Google)
Gemini 3.8 Flash drops the numeric thinking_budget parameter for a three step string enum, thinking_level, and the model errors out if you ask it for the minimal setting that older Gemini versions supported.
Read the tip →Friday, September 11, 2026 · North-Small-Translate-1.0 (Cohere Labs)
Cohere Labs just released an open weight, 218 billion parameter translation model with a built in multi pass workflow that can meaningfully change output quality depending on how you call it. Here is what to test before you decide whether it replaces your current translation vendor.
Read the tip →Wednesday, September 9, 2026 · Mercury 2.5 (Inception)
Inception's Mercury 2.5 shipped Monday and drops into any OpenAI compatible client, which is exactly why two of its behaviors get missed. It accepts a reasoning_effort parameter, so its speed is a dial rather than a fixed property, and its optional diffusing stream returns blocks of noisy tokens that refine in place instead of appending left to right. Set the first deliberately. Keep the second away from anything that parses.
Read the tip →Monday, September 7, 2026 · Claude Mythos 5.1 (Anthropic)
Claude Fable 5.1 invalidates a thinking block the moment you edit an earlier turn, rebuild your system prompt, or compact a conversation. Claude Mythos 5.1 does not run that check at all. If your agent loop mutates history, that difference is the whole migration.
Read the tip →Saturday, September 5, 2026 · MAI-Transcribe-2 (Microsoft)
Microsoft's new transcription model is ten cents an hour, but the Azure AI Speech defaults do not give you it. enhancedMode.enabled is false, diarization is false, and word timestamps are none, so an unmodified request returns a cheap transcript from a different model with none of the structure you wanted.
Read the tip →Friday, September 4, 2026 · GPT-6 Astra
OpenAI's newest model has a 1,050,000 token context window with a pricing cliff two thirds of the way in, and a reasoning effort scale that gained two levels at the top and lost one at the bottom. Both defaults will quietly cost you money if nobody changes them.
Read the tip →Wednesday, September 2, 2026 · Claude Fable 5.1
Two changes in the September 1 release will hit your agent stack on the same day. Forced tool use is gone and now errors out, while cache reads dropped to a quarter of their old price. One breaks your orchestration layer, the other quietly cuts your bill. Handle them together.
Read the tip →Friday, August 28, 2026 · Gemini 3.5 Transcribe
Google's newest speech model ships with a documented tradeoff most teams will switch on without reading: turning on word level timestamps degrades transcription accuracy. Here is what the specification actually allows, and how to configure it so you do not pay for precision you did not need.
Read the tip →Wednesday, August 26, 2026 · GPT-5.6
GPT-5.6 no longer downscales images before it looks at them. Because the detail parameter still defaults to auto, vision code written for older models now sends full resolution images and pays full resolution token prices, with nobody having changed a line.
Read the tip →Monday, August 24, 2026 · Claude Code (Anthropic)
Anthropic made Auto Mode the default in Claude Code on August 14 for Pro, Max, and Team plans, and it comes with hard coded fallback thresholds that determine when the assistant stops asking for permission and when it silently reverts to asking for every action again.
Read the tip →Friday, August 21, 2026 · Gemini 3.7 Flash
Google's own migration checklist for gemini-3.7-flash tells you to delete three sampling parameters most teams have carried in their generation config for years. If you migrated by swapping the model string, you are shipping dead parameters and paying for a thinking level you never chose.
Read the tip →Wednesday, August 19, 2026 · GLM-5.3
Z.ai's GLM-5.3, released August 14, drops the ability to turn reasoning off entirely. Its documentation states that the model supports reasoning only and that disabling reasoning is not supported. The one lever left is reasoning_effort, which accepts low, high and max, and defaults to max. If you port a GLM-5.2 integration without touching it, every call you make, including trivial ones, now runs at the most expensive setting available.
Read the tip →Monday, August 17, 2026 · NVIDIA Nemotron 3.5 Lightning 30B (BF16)
NVIDIA's new Nemotron 3.5 Lightning 30B advertises a one million token context, but the BF16 checkpoint only delivers the full window across eight H100 GPUs with tensor and expert parallelism. On a single 80GB H100 you are validated to 256K, so plan your hardware around the context you actually need before you promise a million.
Read the tip →Friday, August 14, 2026 · Grok 4.6
xAI's new flagship exposes four reasoning effort tiers and quietly defaults to high. Any integration ported over without touching the request payload is now paying for deep reasoning on classification and routing calls that never needed it. Here is how to check, and what the new xhigh tier does to your benchmarks.
Read the tip →Wednesday, August 12, 2026 · Muse Glimmer 30B
Meta's new 30 billion parameter open weight agent fits on a single consumer GPU, and its DFlash speculative decoding drafter is the difference between a demo and a daily driver. Here is how to deploy it properly in its first week.
Read the tip →Monday, August 10, 2026 · Qwen3.8-Max
Alibaba's new 2.4 trillion parameter flagship advertises a 1,000,000 token context window, but its documented maximum input length is 991,808 tokens, and 983,616 in thinking mode. Teams sizing long context jobs against the round number will hit a wall. The bigger lever is the cache: implicit cached input is priced at a quarter of standard input.
Read the tip →Friday, August 7, 2026 · LFM2.5-2.6B
Liquid AI's LFM2.5-2.6B, released 4 August, runs a full tool calling agent at 220 tokens per second on an M5 Max in under 2.5 GB of memory. The move that pays is not replacing your frontier model. It is splitting your agent loop so the high volume mechanical turns run locally at zero marginal cost and only the hard turns hit the paid API.
Read the tip →Wednesday, August 5, 2026 · DeepSeek V4-Flash
Teams adopting DeepSeek V4-Flash 0731 for its fourteen cent input price are quietly paying for high effort reasoning on every call, because thinking mode is on by default. Four sampling parameters they carried over from another provider are being accepted and silently ignored at the same time.
Read the tip →Monday, August 3, 2026 · GPT-5.6
Most teams rebuild GPT-5.6's reasoning from scratch on every turn of a conversation. The model can persist and reuse reasoning items across turns instead, which raises multi turn quality and lifts cache hit rates at the same time.
Read the tip →Friday, July 31, 2026 · Claude Opus 5
Most teams set the effort parameter once and treat it as a thinking length control. It also caps how many tool calls the model makes inside an agentic loop, which is where runaway cost actually comes from.
Read the tip →Wednesday, July 29, 2026 · Claude Opus 5
Asking a frontier model “is this any good?” returns its default taste. Hand it a rubric you wrote and the same model returns consistent, comparable, defensible scoring.
Read the tip →