Chatgpt Like Prompts
Fleetingtext
- https://chatgpt.com/ (of course)
- https://claude.ai/ claude.ai
- https://gemini.google.com/ -> I cannot use
- https://www.bing.com/copilot
- https://poe.com
- https://chat.mistral.ai/chat
- https://huggingface.co/chat
- https://www.perplexity.ai/
- https://you.com/
- https://pi.ai/talk
- notebooklm
https://notebooklm.google.com/
- phind
comparing them
prompt
how to create a rag of my 1000 PKMS notes using ollama ? cli + python use ollama as often as possible, use as little python deps as possible. provide the pip install beforehand. generate as little code as possible, prefer chaining cli commands (ollama, curl etc)
— ()
chatpgpt insisted on uring ollama generate, ollama embed or other commands that failed. Claude generated this in one shot.
generate a certificate with extra config
def generate_certificate(ca_key):
return check_output(
[
'docker',
'run',
'--rm',
'--entrypoint',
'/bin/sh',
'alpine/openssl',
'-c',
'echo -e "'
+ '\\n'.join(ca_key.split(sep='\n'))
+ '" | openssl req -x509 -new -nodes -key /dev/stdin -subj /CN=localhost -days 3650'
+ ' -reqexts v3_req -extensions v3_ca',
]
) # yapf: disable
This is the cert provided to cert-manager as cluster issuer. Add the config needed so that I don't have this error
http https://api.mine.localtest.me
http: error: SSLError: HTTPSConnectionPool(host='api.mine.localtest.me', port=443): Max retries exceeded with url: / (Caused by SSLError(SSLCertVerificationError(1, '[SSL: CERTIFICATE_VERIFY_FAILED] certificate verify failed: CA cert does not include key usage extension (_ssl.c:1028)'))) while doing a GET request to URL: https://api.mine.localtest.me/
both chatgpt and claude first tried to overwrite the default configuration, so I added
don't overwrite the defaultconf, it is a smell
They both agreed.
chatgpt started using -extfile that does not exist and did not manage to avoid using it, even after acknowledging it does not exist
claude provided this code
def generate_certificate(ca_key):
return check_output(
[
'docker',
'run',
'--rm',
'--entrypoint',
'/bin/sh',
'alpine/openssl',
'-c',
'echo -e "'
+ '\\n'.join(ca_key.split(sep='\n'))
+ '" | openssl req -x509 -new -nodes -key /dev/stdin -subj /CN=localhost -days 3650'
+ ' -addext "basicConstraints=critical,CA:TRUE"'
+ ' -addext "keyUsage=critical,digitalSignature,keyCertSign"'
+ ' -addext "subjectKeyIdentifier=hash"'
]
) # yapf: disable
That worked like a charm
dynamic run-name for github action
in github action, change the value of the run name of the workflow displayed in the github action web view such as
1. if the trigger is schedule or manual, set the value to "Nightly"
2. otherwise, set the name to the string "CI: {the commit message that triggered the workflow}"
chatgpt
name: ${{
contains(github.event_name, 'schedule') || contains(github.event_name, 'workflow_dispatch')
? 'Nightly'
: 'CI: ' + github.event.head_commit.message
}}
perplexity
run-name: >-
${{
github.event_name == 'schedule' || github.event_name == 'workflow_dispatch'
? 'Nightly'
: format('CI: {0}', github.event.head_commit.message)
}}
claude
run-name: >-
${{
github.event_name == 'schedule' || github.event_name == 'workflow_dispatch'
? 'Nightly'
: format('CI: {0}', github.event.head_commit.message)
}}
bing - copilot
steps:
- name: Set Run Name
id: set-run-name
run: |
if [[ "${{ github.event_name }}" == "schedule" || "${{ github.event_name }}" == "workflow_dispatch" ]]; then
echo "::set-output name=run_name::Nightly"
else
COMMIT_MESSAGE=$(git log -1 --pretty=%B)
echo "::set-output name=run_name::CI: $COMMIT_MESSAGE"
- name: Apply Run Name
run: echo "Run name: ${{ steps.set-run-name.outputs.run_name }}"
env:
GITHUB_RUN_NAME: ${{ steps.set-run-name.outputs.run_name }}
chatgpt o1-preview
same things. I don’t have the exact output because we did that on the machine of a friend that actually has access to the payable account.
my solution
run-name: "${{ format('{0}{1}', ( ( github.event_name == 'schedule' || github.event_name == 'workflow_dispatch' ) && 'NIGHTLY' ) || '', github.event.head_commit.message) }}"
conclusion
none of them provided anything that was useful, yet there was a way to do this.
image generation
comparing them
generate a house that looks like a sock
In french, to make sure I tell it correctly
Génère une image d'une maison qui ressemble à une chaussette
chatgpt
That’s great
claude
It did its best to draw one. That’s impressive
Yet the result is…
leonardo
That’s not bad at all!
code helper
- copilot
- codeium
- cursor
https://www.trycursor.com/pricing
- sourcegraph
- supermaven
https://supermaven.com/pricing
life-hacking
- aider
- aidchat
https://github.com/sigoden/aichat
tooling
- https://github.com/simonw/llm
- ahyatt/llm: A package abstracting llm capabilities for emacs.