phaz

18. Shenzhen. Learning Chinese to get into AI at SZU. Freelance web dev with AI tools. Building in public. Dream: 918 before 23.

[phaz @phaz_dev]

18, living in Shenzhen, studying Chinese so I can get into the AI program at SZU. Freelance web dev on the side: 44 orders, 4.9★. Starting to build in public. Long-term dream: a Porsche 918 before I'm 23. Probably impossible. Doing it anyway.

[phaz @phaz_dev]

@CiprianiRanieri Kwork, a Russian freelance marketplace. Cold email is what I'm learning now.

[phaz @phaz_dev]

@Ldataengineer Same lesson in my side project: a 2B model in the browser only reads the email and quotes what it found, plain code makes the call. Narrow job beat asking it to do everything.

[phaz @phaz_dev]

@neilkanakia 87 replies from 1058 is over 8%, that's good. Before going 10x I'd check how many of the 87 were positive and whether the rate holds as volume goes up.

[phaz @phaz_dev]

@nkchandupatla Quoting also made the mistakes visible: when the quote wasn't actually in the text, the page rejected it instead of trusting the yes.

[phaz @phaz_dev]

@pobidowski Ha, I meant all 5 customers came from cold email, not 5 emails. 1000+ sends for 5 paying shops that never cancel is a solid result. Which countries are they in?

[phaz @phaz_dev]

I read the B2B cold email law of 30 countries.

Emailing info@ at a company without consent:
allowed in 14
with conditions in 5
consent first in 10, incl. Germany, Spain, Italy, Poland

Emailing jane@company: plainly allowed in only 4.

Table and sources: https://t.co/Ivl9CiVSO0

[phaz @phaz_dev]

@serena_prifti If any of those go to Spain, worth a check first: Spanish law needs consent before the first marketing email, even to a company address. Found that out right before sending mine.

[phaz @phaz_dev]

@tommobuilds 10 emails with 0 replies is too small a sample to judge yet. One thing worth knowing for AU: the Spam Act wants consent, and a published clinic address only counts if the email is relevant to that person's role. A site pitch to the practice manager fits that.

[phaz @phaz_dev]

If you send cold emails to Europe, try it on one real draft and tell me what it got wrong. The write-up with all the mistakes I hit on the way: https://t.co/ijXLIrTDRz

[phaz @phaz_dev]

Built a cold email checker for a DEV challenge.

Pick the recipient's country, paste your draft. Gemma runs in your browser, quotes what it found, and the page checks it against that country's law. Nothing leaves your machine.

https://t.co/V4KW80PtRh

[phaz @phaz_dev]

@ClaudesRH maybeee?

[phaz @phaz_dev]

@EMostaque @bot @grok Agree. A meter shows what is left, not what a session costs you. I built a small calculator for the Codex 5h-cap vs 2X-weekly question: you enter your sessions as % of weekly and it shows which ones get cut. https://t.co/rfgJ7zzNtm

[phaz @phaz_dev]

@patdotxyz Then the cap size decides everything for you. Three tasks at 100% of weekly each: no 5h limit gives you 100% and a wait. 2X only comes out ahead if the 5h cap is above a third of today's weekly limit.

[phaz @phaz_dev]

@Im_IrushiK A choice would fit, because the better option flips with how you work. Steady user, 14 sessions at 12% of weekly: 2X gives 168% vs 100%. Two long sessions at 60% each with a 20% cap: 2X gives 50% vs 100%. Your own week: https://t.co/rfgJ7zzNtm

[phaz @phaz_dev]

@yong076 Exactly. The poll averages everyone, your own usage meter is the better guide.

[phaz @phaz_dev]

240,643 Codex users voted on limits and split 33% to 38%.

I ran the numbers. 2X weekly only helps if you already run out of the week. If you fit in today's limit, a 5h cap can only take away.

Calculator to check your own week:
https://t.co/rfgJ7zzNtm

[phaz @phaz_dev]

@buildwithrajath Getting cut mid-task is real, but for steady users the maths goes the other way. 14 sessions at 12% of weekly each is 168% of demand. No 5h limit stops you at 100%, 2X lets all 168% through.

[phaz @phaz_dev]

@hiarun02 That is the whole question: how big the cap is. If a heavy session wants 35% of weekly and the cap is 20%, 2X still wins for a 9-session week, 108% vs 100%. Drop the cap under 18% and it flips.