SPONSORED
Run open-source LLMs in real production
Capture live traffic, fine-tune and optimize, then deploy your own checkpoints to dedicated GPU endpoints. Choose hardware, set scaling limits, and select region. Stable latency, predictable cost, clear data residency.
SIGNAL PREVIEW
You're building a file mess in ChatGPT. Now there's a Library to prove it
📡 ChatGPT can now find your old files — if you named them well →
🔍 Chrome and Edge Copilot can compare your open tabs →
🛠️ Get the checklist before ChatGPT grabs the wrong file →
Jump to the one that matters to you ↓
📊 This Week's Poll
What frustrates you most about ChatGPT right now?
THE SIGNAL
The file is easier to find now, but the label still matters.
📡 ChatGPT is getting a file drawer
What's new: On May 14, 2026, OpenAI added File Library and Recent files to ChatGPT on web, iOS, and Android. Every document, spreadsheet, or PDF you upload — or that ChatGPT creates for you — is now saved and accessible across chats.
Open a new chat, click the + button, and select Recent files to pull from your last three uploads. Older files live in Library.
You can also ask conversationally: "What should I know about the document I uploaded yesterday?" ChatGPT will find it.
Storage by plan: Free gets 500 MB, Plus and Business get 20 GB, Pro gets 100 GB. Individual files cap at 512 MB with a 2-million-token limit for text documents.
One exception: Temporary Chat. Nothing from a Temporary Chat gets saved — full stop.
ChatGPT is becoming a workspace, not just a chatbot. The danger is the same one you already have with Google Docs and Slack — open a thing, trust the nearest filename, paste it into real work, move on.
That habit breaks here.
The test: I uploaded a PDF brief, a spreadsheet, and a writing guideline, then opened a fresh chat the next day. ChatGPT found the correct file without re-uploading.
Failures showed up fast: Temporary Chat saved nothing, vague filenames pulled the wrong version, and a large PDF lost context halfway through the document.
Want to test this on your own files?
I made the checklist I used to see when ChatGPT reused the right file, picked the wrong version, and failed in Temporary Chat.
DEEPER CUT
Read All Your Open Tabs to Summarize, Compare and Plan Tasks
🔍 Your open tabs are already a research tool
Most people open 10 research tabs, skim all of them manually, then ask ChatGPT for "a summary." That's the lazy part.
Your tabs are already context. In May 2026, both Gemini in Chrome and Copilot in Edge work across multiple open tabs if you allow access. No new AI subscription needed if you already use Chrome or Edge.
Chrome: open the Gemini side panel, share multiple tabs, then ask. Edge: click Copilot, allow tab access, then ask across what's open.
Three prompts worth saving:
"Turn these open tabs into a 5-bullet brief I can send to my manager."
"Compare these options and tell me what I would regret choosing later."
"List where these pages disagree, and quote the line from each page."The prompt that changed the output wasn't "summarize this." It was the comparison and disagreement one.
Both systems handled basic summaries in testing (May 2026, Chrome and Edge). Both failed when pushed past 5 dense tabs — contradictions got smoothed into one bland consensus instead of flagged clearly.
That matters more than the summary itself.
If money, contracts, health, or code are involved: skim the originals yourself. AI still defaults toward sounding coherent over sounding careful.
🛠️ Use This Tuesday
🛠️ Reuse a file already inside ChatGPT — without opening Drive, Finder, or downloads
What it does: Pull a file from ChatGPT's Library into a new chat in under 60 seconds.
Pull once & reuse
How to do it:
Open a new ChatGPT chat
Click the + button in the message box
Select Recent files
Pick one of the last 3 files shown
Ask: "Use this file and tell me the 5 things I need to remember before I work on this again."
If the file is older: click Add from library and search for it by name
One common mistake: Don't ask ChatGPT to "use my file" when you have multiple versions named things like "Brief_v2" or "Final_Brief." That's how you quietly summarize the wrong document.
Expected output: A working summary from an already-uploaded file in under 60 seconds — no re-upload, no searching your downloads folder.
Or reply REUSE and I'll send it directly.
Going Deeper: If you reuse files across multiple projects, ask ChatGPT to name the exact file, page, sheet, or row before it answers anything.
📨 Forward this to the person who uploads the same file to ChatGPT every Monday.
From the Community
📬 From the Community
Last week's poll asked: what would make AI Signal more useful to you?
🟩🟩🟩🟩🟩🟩 🛠️ More step-by-step tutorials (65)
🟨🟨🟨⬜️⬜️⬜️ 🧠 More honest tool comparisons (33)
🟨⬜️⬜️⬜️⬜️⬜️ 💼 More AI career and business impact (21)
119 Votes
Step-by-step won by 28 points. The replies backed it up.
Linda N. saves every sample prompt into a cheat file she reuses.
Pamela C. put it plainly: understand HOW before you compare tools.
That's the mandate this issue follows.
Doreen K. uses only Copilot at work and wants more Copilot walkthroughs specifically.
She wasn't alone — I'm paying attention.
How was today's issue?
AI tools are getting better at seeing the stuff already on your screen — old files, open tabs, prior context — but you still have to check what the AI is actually using.
📨 Forward this to one person whose ChatGPT uploads folder is already a junk drawer.
Hit reply and tell me the one file you keep re-uploading to ChatGPT.
Dan Rice · AI Signal · Every Tuesday