From my initial testings GPT-5 is as bad as GPT-4o on instruction following. This is a real bummer. No improvement at all – but I need more testing to validate this.
PROMPT ENGINEERING
-
Are LLM Factual Errors Hallucinations? A Definitional Question
By
–
Follow-up question: does "there are three Bs in blueberry" count as a hallucination? (I expect these poll numbers will be much closer to each other)
-
Router overconfidence limits question routing to simpler models
By
–
El problema es que el enrutador sobre-confía en la dificultad de las preguntas, tirándolas al modelo más simple y fallando. En esta pregunta no ha querido razonar y GPT-5 se ha equivocado. Pero si se la mando directamente al GPT-5 Thinking, sólo pensando 10 segundo, acierta.
-
IA Summer Lessons: Your Essential AI Action Checklist
By
–
𝐋𝐞ç𝐨𝐧 𝟓- 𝐄𝐩𝐢𝐬𝐨𝐝𝐞 𝟑 : 𝐕𝐨𝐭𝐫𝐞 𝐜𝐡𝐞𝐜𝐤𝐥𝐢𝐬𝐭 𝐈𝐀 𝐀𝐜𝐭 Pas question de partir à l’aventure sans ton panier pique-nique réglementaire ! Bienvenue dans mon #FlashCahierDeVacances – IA Edition Le seul programme d’été pour tenir la barre et ne pas
-
Testing LLM Math Problem Solving and Simple Explanations
By
–
Here’s the test prompt we used: "A small bakery sells bread for $4 each. It costs them $2 to make a loaf. They have fixed costs of $600 a month. They want to earn $1,000 profit in a month. How many loaves do they need to sell? Also, give the answer in a way a 12-year-old could
-
GPT-5 Generates Fully Functional Multi-Page Websites Single Prompt
By
–
10. GPT-5 creates a fully functioning multi-page website from just a single prompt.pic.twitter.com/9dcU8qAQmQ
— Shubham Saboo (@Saboo_Shubham_) 8 août 202510. GPT-5 creates a fully functioning multi-page website from just a single prompt.
-
GPT-5 Solves Hexagon Prompt Challenge Successfully
By
–
7. GPT-5 gets the ball in Hexagon prompt in one go.pic.twitter.com/Yuut7jzE1s
— Shubham Saboo (@Saboo_Shubham_) 8 août 20257. GPT-5 gets the ball in Hexagon prompt in one go.
-
Realistic Minecraft Clone Generated with Single AI Prompt
By
–
5. A realistic Minecraft clone with one simple prompt.pic.twitter.com/LYamQPaPNz
— Shubham Saboo (@Saboo_Shubham_) 8 août 20255. A realistic Minecraft clone with one simple prompt.
-
GPT-5 Launch: 10 Insane Use Cases Already Built
By
–
GPT-5 just dropped less than 9 hours ago.
— Shubham Saboo (@Saboo_Shubham_) 8 août 2025
And people are already building wild stuff with it.
Here are 10 insane use cases I’ve seen so far:
1. Snake game in a Hexagon pic.twitter.com/qfws7RvFX2GPT-5 just dropped less than 9 hours ago. And people are already building wild stuff with it. Here are 10 insane use cases I’ve seen so far: 1. Snake game in a Hexagon
-
GPT-5 Launch Party: Search Improvements and Context Window Priorities
By
–
livetweeting launch party “gpt5 is much better at search btw” they didnt get to emphasize everyone voted for 10m context window over better RFT