Instagram-Story-Viewer> @itstundeanthony> Beiträge
305
Beiträge
198.3K
Follower
790
Folge ich
I talk about building apps
YCombinator | Ex Google
[email protected]
Find the best tech stack for your app ⤵️
BEITRÄGE STORIES REELS MARKIERT
Alle Herunterladen
Wait for assembly 😮#programming #code by @itstundeanthony
3k
vor 2 Jahren
Herunterladen
Comment “code” to get the source code #programming #code by @itstundeanthony
1k
vor 2 Jahren
Herunterladen
Comment “eval” to get a link to this library#braintrustpartner #code by @itstundeanthony
23
vor 6 Stunden
Herunterladen
It’s better to start again sometimes 😵 by @itstundeanthony
69
vor 4 Tagen
Herunterladen
comment “yes” to get Readdy AI that can redesign your entire website#readypartner #code #programming by @itstundeanthony
57
vor 6 Tagen
Herunterladen
Is coding still fun? by @itstundeanthony
42
vor 11 Tagen
Herunterladen
This one could get you hacked 👀#programming #code by @itstundeanthony
27
vor 12 Tagen
Herunterladen
Making an app inspired by… by @itstundeanthony
11
vor einem Monat
Herunterladen
For my non technical friends 🫶 Comment “guide” to get a free guide on this#emergentpartner #programming #code by @itstundeanthony
217
vor einem Monat
Herunterladen
I wish I knew about this earlier 😍! The library I’m using is called DiceBear, it’s open source and has thirty different styles for free.Leave a “yes” if you’ll try this in your future projects 🙌🏽#programming #code by @itstundeanthony
31
vor einem Monat
Herunterladen
I spend 80% of my time making a very detailed plan before I write even one line of code, because a good plan saves a lot of time and tokens along the way and minimizes the back-and-forth with AI.Comment “template” to get a link to the miro template I use#miropartner #programming by @itstundeanthony
58
vor einem Monat
Herunterladen
Here’s how I brought the cost down 👇🏽I’m building an inbox management AI app, and every time a user runs it, the AI has to read their entire inbox (Gmail or Instagram). Here’s how I cut the cost per run:1. Prompt caching: Anthropic and OpenAI both let you cache large context for a 90% discount on input tokens. So I started caching the inbox between calls.2. Switching to a cheaper model: I built an eval to test model quality and found that Sonnet and Haiku produced the same output for my app, so I switched to Haiku, which is waaay cheaper.3. Removing tools: Every time an agent uses a tool, your costs multiply. In the early version of my app, I gave the model a bunch of tools to search the inbox, which got expensive fast. The fix: I removed the tools and passed the inbox directly into the context.Drop a “Yes” in the comments if you found this useful.I’m building a new app every month and documenting what I learn along the way.#programming #code by @itstundeanthony
31
vor einem Monat
Herunterladen
×

Download all media on this page

Photos Videos
back to up