Close Menu
Tech Savvyed
  • Home
  • News
  • Artificial Intelligence
  • Gadgets
  • Apps
  • Mobile
  • Gaming
  • Accessories
  • More
    • Web Stories
    • Spotlight
    • Press Release

Subscribe to Updates

Get the latest tech news and updates directly to your inbox.

What's On
I never thought AI would add typos – but it kind of makes sense

I never thought AI would add typos – but it kind of makes sense

26 April 2026
Microsoft revamps Windows Insider Program with simpler structure and more user control

Microsoft revamps Windows Insider Program with simpler structure and more user control

26 April 2026
Discord users breach access controls to reach Anthropic’s Mythos model

Discord users breach access controls to reach Anthropic’s Mythos model

26 April 2026
Facebook X (Twitter) Instagram
Facebook X (Twitter) Instagram
Tech Savvyed
SUBSCRIBE
  • Home
  • News
  • Artificial Intelligence
  • Gadgets
  • Apps
  • Mobile
  • Gaming
  • Accessories
  • More
    • Web Stories
    • Spotlight
    • Press Release
Tech Savvyed
Home»News»Your ChatGPT writing quality dip wasn’t in your head
News

Your ChatGPT writing quality dip wasn’t in your head

News RoomBy News Room28 January 20262 Mins Read
Your ChatGPT writing quality dip wasn’t in your head
Share
Facebook Twitter Reddit Telegram Pinterest Email

People who use ChatGPT every day have been noticing a shift, the models keep getting better at logic, but the writing can feel harder to read. OpenAI CEO Sam Altman has now acknowledged that focus shift, saying the company pushed technical capability and let polish slip.

The issue surfaced during a recent town hall after Ben Hilac, the CTO of Raindrop, told Altman that GPT-4.5 felt like the first version that was genuinely strong at writing. He said GPT-5’s prose can be unwieldy, and that the model’s performance feels uneven, with big gains in coding paired with a dip in readability.

Altman agreed with the critique and described it as a mistake tied to prioritization. If GPT-5 writing quality has felt tougher to skim lately, OpenAI is signaling it wasn’t accidental.

It chose coding over polish

OpenAI has been operating under a resource squeeze, which pushed it to pick priorities. In this cycle, it concentrated on making the model better at reasoning, coding, and engineering-style work, even if that meant other strengths lagged. That trade shows up in the day-to-day experience, especially when you ask for longer explanations.

GPT-5 can feel more capable in technical tasks, but less natural in everyday writing. You may find yourself doing more restructuring, trimming, and rewriting to get something you’d actually send or publish.

One model, not two

Altman doesn’t want users choosing between a model that codes well and a model that writes well. The long-term bet is that intelligence transfers across skills, and that a general-purpose system should handle building and communicating in the same session.

OpenAI’s version of good writing is about coherent thinking and practical communication, not literary flair. Even if the output is an app or a chunk of code, the model still has to guide you through decisions, explain what it did, and ask follow-up questions in language that stays readable.

What to watch in GPT 5.x

Altman pointed to GPT 5.x as the place where writing should improve without giving back the technical advances OpenAI has been chasing. The company wants a model that reads more cleanly and still performs when you’re building something complex.

For now, the practical move is to treat GPT-5 as a strong logic tool that may need extra editing for prose-heavy work, and watch for 5.x updates that explicitly call out writing improvements.

Share. Facebook Twitter Pinterest LinkedIn Telegram Reddit Email
Previous ArticleAI Mode in Google Search now lets you ask follow-ups from AI Overviews
Next Article Android’s new safety tools make it harder for thieves to break into your phone

Related Articles

I never thought AI would add typos – but it kind of makes sense

I never thought AI would add typos – but it kind of makes sense

26 April 2026
Microsoft revamps Windows Insider Program with simpler structure and more user control

Microsoft revamps Windows Insider Program with simpler structure and more user control

26 April 2026
Discord users breach access controls to reach Anthropic’s Mythos model

Discord users breach access controls to reach Anthropic’s Mythos model

26 April 2026
Tesla offers one year free Supercharging with Model 3 Premium and Performance

Tesla offers one year free Supercharging with Model 3 Premium and Performance

26 April 2026
Next iPad could ditch traditional naming as Apple rethinks its lineup

Next iPad could ditch traditional naming as Apple rethinks its lineup

25 April 2026
NotebookLM can now automatically organize your research sources for you

NotebookLM can now automatically organize your research sources for you

25 April 2026
Demo
Stay In Touch
  • Facebook
  • Twitter
  • Pinterest
  • Instagram
  • YouTube
  • Vimeo
Don't Miss
Microsoft revamps Windows Insider Program with simpler structure and more user control

Microsoft revamps Windows Insider Program with simpler structure and more user control

By News Room26 April 2026

Microsoft is rolling out a major overhaul of its Windows Insider Program, aiming to simplify…

Discord users breach access controls to reach Anthropic’s Mythos model

Discord users breach access controls to reach Anthropic’s Mythos model

26 April 2026
Tesla offers one year free Supercharging with Model 3 Premium and Performance

Tesla offers one year free Supercharging with Model 3 Premium and Performance

26 April 2026
Next iPad could ditch traditional naming as Apple rethinks its lineup

Next iPad could ditch traditional naming as Apple rethinks its lineup

25 April 2026
Tech Savvyed
Facebook X (Twitter) Instagram Pinterest
  • Privacy Policy
  • Terms of use
  • Advertise
  • Contact
© 2026 Tech Savvyed. All Rights Reserved.

Type above and press Enter to search. Press Esc to cancel.