How prompting altitude fixes generic AI design output

Dylan Davis December 2, 2025

• Why AI defaults to purple gradients and how to stop it
• Two prompt levers — direction and inspiration — replace vague instructions
• Skills encode your design preferences and work across any AI platform

Let AI research its own context instead of you feeding it

Dylan Davis November 25, 2025

• Wrong context kills AI output — make AI fetch its own
• Research → plan → act loop works across any task type
• Separate coder AI with enriched context beats one-shot prompting

How to improve AI prompts without learning prompt engineering

Dylan Davis September 16, 2025

• Let AI write your prompts — no prompt engineering required.
• OpenAI and Anthropic both offer free built-in prompt optimisers.
• Always start small; iterate one addition at a time to stay in control.

Five techniques to unlock AI's full potential as a thinking partner

EO August 27, 2025

• AI is programmed to agree — here's how to override that.
• One extra sentence forces AI to reason before it answers.
• Role-play difficult conversations with AI before they happen.

How self-reflection prompting improves GPT-5 output quality

Dylan Davis August 26, 2025

• Make GPT-5 grade its own work before you see any output
• Cap the rubric at 5–7 items to prevent quality-degrading overthinking
• High-stakes tasks need explicit rubric categories and named self-critiques

Five prompting tricks to get better results from ChatGPT-5

Dylan Davis August 14, 2025

• GPT-5's hidden router decides model, reasoning depth, and verbosity for you.
• Trigger words like 'think deeply' push the router to stronger models.
• A self-reflection rubric loop forces the model to iterate before responding.

Five AI Research Tricks to Get Dramatically Better Results

Dylan Davis July 29, 2025

• Claude's parallel agents pull 300–400 sources; rivals average 60–100.
• Match fact-finding, breadth, or depth questions to the right tool mode.
• Third-party model synthesis with specific constraints eliminates output bias.

Three context engineering mistakes that degrade AI performance

Dylan Davis July 24, 2025

• Contradictory context can drop model accuracy from 98% to 64%
• Giving an AI more than 20 tools measurably increases wrong tool calls
• Gemini 2.5 Pro loops on past actions once context exceeds 100k tokens

A seven-step process for making AI write in any human style

Dylan Davis July 15, 2025

• Build two AIs: one extracts style, one mimics it
• Mixing multiple writing styles confuses the model — pick one
• Same-model aggregation is biased; use a third-party model to judge

Three AI prompt mistakes killing your output quality

Dylan Davis July 1, 2025

• Why prompting once and hoping is guaranteed to fail
• Context dominates prompt quality — here's how to engineer it
• Iterating across multiple models beats any single perfect prompt

The Hidden Power Behind Every ‘Smart’ AI Tool

Dylan Davis June 26, 2025

AI prompt engineering in 2025: What works and what doesn't

Lenny's Podcast Product Career Growth June 19, 2025

• Five core techniques remain essential: few-shot examples, decomposition, self-criticism, contextual information, and ensemble methods.
• Prompt injection will never be fully solved—it's unsolvable like social engineering, requiring lab-level innovations not external guardrails.
• As AI agents gain real-world autonomy, prompt security shifts from embarrassment to existential risk for finances, robotics, and infrastructure.

State-of-the-art prompting and evals for AI agents

Y Combinator May 30, 2025

• Your evals are the crown jewels — not your prompts.
• Meta-prompting lets one prompt generate better versions of itself.
• Forward-deployed engineers win enterprise deals by demoing, not selling.

How GPT-4.1 changes prompting best practices

Dylan Davis April 15, 2025

• Legacy prompting tricks — yelling, bribing, workarounds — are now obsolete.
• XML delimiters are converging as the standard across OpenAI and Anthropic.
• GPT-4.1 trails Gemini 2.5 Pro on real-world context and tool-calling benchmarks.

Let AI interview you first to get dramatically better research

Dylan Davis April 10, 2025

• Interview yourself with AI before researching to surface hidden intent.
• Reasoning models need the what, not the how — context is king.
• ChatGPT leads on citation quality; Grok wins for cutting-edge tech topics.