Mo

396 posts

Mo banner
Mo

Mo

@heyitsmo_1

Building in public, Sharing tools and knowledge.

Katılım Eylül 2022
349 Takip Edilen532 Takipçiler
Mo
Mo@heyitsmo_1·
𝘄𝗮𝗻𝘁 𝘁𝗼 𝗯𝘂𝗶𝗹𝗱 𝘄𝗶𝘁𝗵 𝗹𝗮𝗿𝗴𝗲 𝗹𝗮𝗻𝗴𝘂𝗮𝗴𝗲 𝗺𝗼𝗱𝗲𝗹𝗦? 𝗽𝗿𝗲𝗽𝗮𝗿𝗲 𝘁𝗼 𝗽𝗮𝘆. these models require 𝗲𝗻𝗼𝗿𝗺𝗼𝘂𝘀 𝗰𝗼𝗺𝗽𝘂𝘁𝗶𝗻𝗴 𝗽𝗼𝘄𝗲𝗿 and 𝗲𝗹𝗲𝗰𝘁𝗿𝗶𝗰𝗶𝘁𝘆. the 𝗰𝗼𝗺𝗽𝘂𝘁𝗮𝘁𝗶𝗼𝗻𝗮𝗹 𝗰𝗼𝘀𝘁 is significant. the 𝗳𝗶𝗻𝗮𝗻𝗰𝗶𝗮𝗹 𝗰𝗼𝘀𝘁 is significant. 𝗧𝗵𝗶𝘀 𝗿𝗲𝘀𝗼𝘂𝗿𝗰𝗲 𝗶𝗻𝘁𝗲𝗻𝘀𝗶𝘁𝘆 𝗰𝗿𝗲𝗮𝘁𝗲𝘀 𝗮 𝗯𝗮𝗿𝗿𝗶𝗲𝗿 𝘁𝗼 𝗲𝗻𝘁𝗿𝘆 𝗳𝗼𝗿 𝗺𝗼𝘀𝘁 𝗯𝘂𝗶𝗹𝗱𝗲𝗿𝘀.
English
0
0
0
12
Mo
Mo@heyitsmo_1·
@Sar_IFO @sudopimp currently am doing a series about prompt engineering follow me for more x.com/i/status/20786…
Mo@heyitsmo_1

𝘆𝗼𝘂 𝗱𝗼𝗻'𝘁 𝗻𝗲𝗲𝗱 𝘁𝗼 𝗯𝗲 𝘁𝗲𝗰𝗵𝗻𝗶𝗰𝗮𝗹 𝘁𝗼 𝗲𝗻𝗴𝗶𝗻𝗲𝗲𝗿 𝗽𝗿𝗼𝗺𝗽𝘁𝘀. prompt engineering is a real discipline. it needs 3 things: 1. 𝗗𝗼𝗺𝗮𝗶𝗻 𝗞𝗻𝗼𝘄𝗹𝗲𝗱𝗴𝗲 - understand your problem deeply 2. 𝗠𝗼𝗱𝗲𝗹 𝗨𝗻𝗱𝗲𝗿𝘀𝘁𝗮𝗻𝗱𝗶𝗻𝗴 - know what the ai can and can't do 3. 𝗠𝗲𝘁𝗵𝗼𝗱𝗶𝗰𝗮𝗹 𝗔𝗽𝗽𝗿𝗼𝗮𝗰𝗵 - test, iterate, refine (like version control in software) example: a lawyer writing legal document prompts needs domain knowledge (contracts), model understanding (what language models miss), and method (testing different instructions to get consistent output). that's engineering. not guessing.

English
0
0
0
6
Mo
Mo@heyitsmo_1·
Me to Claude: "Make no errors. DO NOT HALLUCINATE. YOU ARE AN SOFTWARE ENGINEER EXPERT"
English
4
0
34
3K
Mugen
Mugen@joseph_mugen·
@heyitsmo_1 Claude: "I'm being spicy right now"
English
1
0
1
24
Mo
Mo@heyitsmo_1·
𝘆𝗼𝘂𝗿 𝗴𝗲𝗻𝗲𝗿𝗮𝗹 𝗽𝘂𝗿𝗽𝗼𝘀𝗲 𝗔𝗜 𝗶𝘀 𝗻𝗼𝘁 𝗯𝘂𝗶𝗹𝘁 𝗳𝗼𝗿 𝘆𝗼𝘂𝗿 𝗶𝗻𝗱𝘂𝘀𝘁𝗿𝘆. 𝗟𝗮𝗻𝗴𝘂𝗮𝗴𝗲 𝗺𝗼𝗱𝗲𝗹𝘀 𝗮𝗿𝗲 𝗴𝗲𝗻𝗲𝗿𝗮𝗹𝗶𝘀𝘁 𝗯𝘆 𝗻𝗮𝘁𝘂𝗿𝗲. they excel at common tasks across all domains. but when you need them in 𝘀𝗽𝗲𝗰𝗶𝗮𝗹𝗶𝘇𝗲𝗱 𝗳𝗶𝗲𝗹𝗱𝘀, 𝘁𝗵𝗲𝘆 𝘀𝘁𝗿𝘂𝗴𝗴𝗹𝗲. to excel in your niche, they need 𝗱𝗼𝗺𝗮𝗶𝗻 𝘀𝗽𝗲𝗰𝗶𝗳𝗶𝗰 𝗱𝗮𝘁𝗮.
English
0
0
0
38
Mo
Mo@heyitsmo_1·
@vishnu_aniyan still not enough some Fields cant accept "pretty close" for example law field.
English
1
0
1
7
Vishnu Aniyan
Vishnu Aniyan@vishnu_aniyan·
@heyitsmo_1 temperature = 0 gets you pretty close to deterministic. not perfect, but consistent enough for most production use cases that need reliable outputs.
English
1
0
1
8
Mo
Mo@heyitsmo_1·
𝗮𝘀𝗸 𝘆𝗼𝘂𝗿 𝗔𝗜 𝘁𝗵𝗲 𝗲𝘅𝗮𝗰𝘁 𝘀𝗮𝗺𝗲 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻 𝘁𝘄𝗶𝗰𝗲. 𝗚𝗲𝘁 𝘁𝘄𝗼 𝗱𝗶𝗳𝗳𝗲𝗿𝗲𝗻𝘁 𝗮𝗻𝘀𝘄𝗲𝗿𝘀. this is not a bug. 𝗟𝗮𝗻𝗴𝘂𝗮𝗴𝗲 𝗺𝗼𝗱𝗲𝗹𝘀 𝗮𝗿𝗲 𝗽𝗿𝗼𝗯𝗮𝗯𝗶𝗹𝗶𝘀𝘁𝗶𝗰. 𝗧𝗵𝗲𝘆 𝗱𝗼 𝗻𝗼𝘁 𝗴𝗶𝘃𝗲 𝘆𝗼𝘂 𝘁𝗵𝗲 𝘀𝗮𝗺𝗲 𝗼𝘂𝘁𝗽𝘂𝘁 𝗲𝘃𝗲𝗿𝘆 𝘁𝗶𝗺𝗲, even with an identical prompt. this breaks consistency in any application 𝘁𝗵𝗮𝘁 𝗻𝗲𝗲𝗱𝘀 𝗿𝗲𝗹𝗶𝗮𝗯𝗹𝗲 𝗿𝗲𝘀𝗽𝗼𝗻𝘀𝗲𝘀.
English
1
0
0
63
Mo
Mo@heyitsmo_1·
gonna drop a solution to minimize the problem follow along so you don't miss it
English
0
0
0
13
Mo
Mo@heyitsmo_1·
𝘆𝗼𝘂𝗿 𝗔𝗜 𝘀𝗼𝘂𝗻𝗱𝘀 𝗰𝗼𝗻𝗳𝗶𝗱𝗲𝗻𝘁. 𝗶𝘁 𝘀𝗼𝘂𝗻𝗱𝘀 𝗰𝗼𝗿𝗿𝗲𝗰𝘁. 𝗯𝘂𝘁 𝗶𝘁 𝗺𝗮𝗱𝗲 𝘂𝗽 𝘁𝗵𝗲 𝗲𝗻𝘁𝗶𝗿𝗲 𝘁𝗵𝗶𝗻𝗴. 𝗟𝗮𝗻𝗴𝘂𝗮𝗴𝗲 𝗺𝗼𝗱𝗲𝗹𝘀 𝗴𝗲𝗻𝗲𝗿𝗮𝘁𝗲 𝗽𝗹𝗮𝘂𝘀𝗶𝗯𝗹𝗲 𝘆𝗲𝘁 𝗳𝗮𝗰𝘁𝘂𝗮𝗹𝗹𝘆 𝗶𝗻𝗰𝗼𝗿𝗿𝗲𝗰𝘁 𝗶𝗻𝗳𝗼𝗿𝗺𝗮𝘁𝗶𝗼𝗻 𝗮𝗹𝗹 𝘁𝗵𝗲 𝘁𝗶𝗺𝗲. this is called 𝗵𝗮𝗹𝗹𝘂𝗰𝗶𝗻𝗮𝘁𝗶𝗼𝗻. it happens because the model does not actually 𝗸𝗻𝗼𝘄 𝘄𝗵𝗮𝘁 𝗶𝘀 𝘁𝗿𝘂𝗲 𝗼𝗿 𝗳𝗮𝗹𝘀𝗲. it just predicts what words come next.
English
1
0
0
41
Mo
Mo@heyitsmo_1·
@DomJohnnie yep its good idid not try it my self but it looks "Outstanding"
English
0
0
1
14
Mo
Mo@heyitsmo_1·
SOMEONE BUILT A MAC APP THAT PUTS CLAUDE CODE, CODEX, CURSOR, GROK, AND MORE DIRECTLY IN YOUR NOTCH.
English
5
0
12
3K
Mo
Mo@heyitsmo_1·
𝗱𝗼 𝘆𝗼𝘂 𝘁𝗵𝗶𝗻𝗸 𝘆𝗼𝘂𝗿 𝗔𝗜 𝗰𝗵𝗮𝘁𝗯𝗼𝘁 𝗿𝗲𝗺𝗲𝗺𝗯𝗲𝗿𝘀 𝘆𝗼𝘂? it does not. 𝗟𝗮𝗿𝗴𝗲 𝗹𝗮𝗻𝗴𝘂𝗮𝗴𝗲 𝗺𝗼𝗱𝗲𝗹𝘀 𝗵𝗮𝘃𝗲 𝘇𝗲𝗿𝗼 𝗽𝗲𝗿𝘀𝗶𝘀𝘁𝗲𝗻𝘁 𝗺𝗲𝗺𝗼𝗿𝘆. 𝗘𝘃𝗲𝗿𝘆 𝗰𝗼𝗻𝘃𝗲𝗿𝘀𝗮𝘁𝗶𝗼𝗻 𝘀𝘁𝗮𝗿𝘁𝘀 𝗰𝗼𝗺𝗽𝗹𝗲𝘁𝗲𝗹𝘆 𝗳𝗿𝗲𝘀𝗵. when the chatbot acts like it knows you, that is 𝗻𝗼𝘁 𝘁𝗵𝗲 𝗺𝗼𝗱𝗲𝗹 𝗱𝗼𝗶𝗻𝗴 𝘁𝗵𝗲 𝗿𝗲𝗺𝗲𝗺𝗯𝗲𝗿𝗶𝗻𝗴. that is 𝗮𝗻 𝗮𝗽𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻 𝗯𝘂𝗶𝗹𝘁 𝗮𝗿𝗼𝘂𝗻𝗱 𝗶𝘁.
English
1
0
2
154
Mo
Mo@heyitsmo_1·
@abdulazeem_s i always drop tools and useful knowledge follow me for more: x.com/i/status/20786…
Mo@heyitsmo_1

𝗺𝗼𝘀𝘁 𝗽𝗲𝗼𝗽𝗹𝗲 𝘂𝘀𝗲 𝘁𝗵𝗲 𝘄𝗿𝗼𝗻𝗴 𝗽𝗿𝗼𝗺𝗽𝘁 𝘁𝘆𝗽𝗲. here are the 3 that matter: type 1 - 𝗜𝗻𝘀𝘁𝗿𝘂𝗰𝘁𝗶𝗼𝗻𝘀 + 𝗤𝘂𝗲𝘀𝘁𝗶𝗼𝗻 (use when you need advice or guidance.): "how should i write my essay? cover sections, tone, and words to avoid." type 2 - 𝗜𝗻𝘀𝘁𝗿𝘂𝗰𝘁𝗶𝗼𝗻𝘀 + 𝗗𝗮𝘁𝗮 (use when you want the AI to create something from your input.): "write 4 paragraphs from this bio: [your background and experience]." type 3 - 𝗤𝘂𝗲𝘀𝘁𝗶𝗼𝗻 + 𝗘𝘅𝗮𝗺𝗽𝗹𝗲𝘀 (use when you want recommendations or pattern matching.): "i like breaking bad and peaky blinders. hate ted lasso. what should i watch?" match the prompt to the task. better prompts → better results.

English
0
0
1
11
0xMarioNawfal
0xMarioNawfal@RoundtableSpace·
SOMEONE BUILT A MAC APP THAT PUTS CLAUDE CODE, CODEX, CURSOR, GROK, AND MORE DIRECTLY IN YOUR NOTCH.
English
25
6
213
64.1K
Mo
Mo@heyitsmo_1·
am launching a complete series on AI prompt engineering. You will learn actionable strategies to get precise results from AI models. The series covers three main pillars. Fundamentals: Prompt structure, types, and LLM limitations... Intermediate: Chain of Thought, hallucination reduction, special tokens, and self-correction.... Advanced: Tree of Thought, tool integration, reflection loops, workflows, and guardrails... Follow to get every breakdown.
English
0
0
0
23
Mo
Mo@heyitsmo_1·
When Fable thinks for 25 mins and hits your session limit before giving your response.
English
2
0
4
122
Mo
Mo@heyitsmo_1·
POV: The senior engineer, losing it after getting fed up
English
0
1
22
4.5K
AJ / J7K Labs
AJ / J7K Labs@J7K_dev·
@heyitsmo_1 tell codex to snoop in the project folder and see what it was doing lol
English
0
0
1
34
Moniiii
Moniiii@miniii_codes·
We have 500 verified followers now. 🎉 Just one step away from monetization. But honestly, I don't know if I'll ever reach it. Between a full-time job and responsibilities at home, finding enough time to create consistently is tough.
Moniiii tweet media
English
51
1
65
1.7K

Keşfet

@IDFkushoverlord @Sar_IFO @sudopimp @joseph_mugen @vishnu_aniyan @DomJohnnie @DeependraGaur15 @elonmusk