Alan on AI
Region: US
Friday 04 September 2026 13:05:48 GMT
1402
65
95
2
Music
Download
Comments
alfaric1 :
kimi coder 2.7 is solid with fraction cost
2026-09-05 08:48:57
0
Unknown :
I use Base44 which is great for me so far.
2026-09-05 04:21:16
1
Ike :
GPT-6 Astra vs Fable 5.1
2026-09-04 15:27:45
1
USAAF13 :
Im using claude code, but I don’t use the opus 5 model anymore, to many faults,, back to 4.8 which works perfectly.. fable only for new chapters for making the plan / design.. also use opencode with omniroute for “free” AI, but those results vary😅
2026-09-04 17:16:13
1
richardhickman82 :
I loved claude code, but for what i do,I cant go higher that opus 4.8 without hitting limits. I tried 5.1 but destroyed my session in 15 mins! the last job I gave it was set me up for handover to codex when needed. im going to have to swap to codex. I considered upgrading to 20x but its only twice the allowance of my 5x 😳
2026-09-04 14:49:07
2
Everett :
I had an issue with Bolt. but switched to. Lauderdale code. love it.
2026-09-04 13:34:55
1
Jason :
All Claude everything
2026-09-05 00:10:45
1
Rocky :
I think its structure. Claude give some really bad drift if your context become too vague. I use cli on ubuntu. And built a custom ui around that.
2026-09-04 16:34:56
1
Brandon🖤 :
Tried opencode. For an always free cli agent its actually so decent
2026-09-05 10:08:27
1
NelsaAl :
i used deepseek, is the best trust try it atleast once
2026-09-05 04:25:43
1
$ :
So far im weighing more to codex specially to add a feature or fix an issue for an existing app
2026-09-04 17:34:47
1
aceofspanners :
Cursor and Goose, both do a super nice job. Cursor is best at making changes visible and having a conversation at the same time
2026-09-04 18:20:53
1
Pauly Be :
I love antigravity, I've been using it since it came out. its just a cleaner UI
2026-09-04 13:26:06
1
Xyclos :
Me no problem coding with Claude since 7 months ago. Great team member.
2026-09-04 18:23:28
1
Label Me Not :
I’ve noticed a decrease in efficiency. I spend a lot more but the actual output seems to be similar.
2026-09-04 17:39:23
1
Kristaps Chris Dinsbergs :
the next grok models till the grok 5 will be dominating the field. specially on engineering as they are now training the 4.7 grok on the 25 years of precision engineering data from spaceX. claude I stopped a long time ago as it was ridiculously low usage for it I switched to OpenAI codex 200 plan never looked back codex wise and Gpt work wise smooth experience. also a fun fact that you will agree with its mostly about the driver behind the wheel who uses any tools to get the results we seek.💪 so mostly all can run smooth and do all u need as its the harness setup that plays a solid role and I switched from loops to temporal logic steps for the agents that was an massive change forwards efficiency and more robust gates between the steps and workflows. been trying to push my harness for open-source but so busy barely have any time for it these days been working on it sense the very beginning of vibecoding which was for me the googles firebase I think forgot the name been ages😂😂 also YOU SHOULD DEFINITELY LOOK IN TO "GROK BOT" this is something massive.💪
2026-09-04 14:33:40
1
jk.20246 :
Wall of text is the absolute worst change
2026-09-04 18:52:29
1
mayford jones :
I use Claude. I paid for it so I’m not moving away.
2026-09-04 14:57:24
1
Shoes Optional Misadventures :
For whatever reason, those of us in the government have been given a plethora to try and by far, for us anyway Claude is it!!! my developer uses it for coding and I use it in design but also to create stories and test steps!!! I’ve tried using the other models, but at least for what we do, no one compares
2026-09-04 21:47:17
1
Retireinspire :
Like claude but I feel,like they are not giving the right amount of credit for each level
2026-09-04 18:38:55
1
Agoragentic :
Stopped using Claude when it put its watermark in my code
2026-09-04 17:45:12
1
Milkywaygrad :
Small easy tasks then directly to Claude code. Complex, Gpt prompting, code following directions, then back me directing gpt for the next prompt. For complex tasks without receiving help for the prompts but directly giving them to Claude made things very messy. Now it's all perfect. Maybe it's my fault, don't know. Found my peace even if it needs 4x more time.
2026-09-04 20:42:02
1
landman :
opus 5 finds fake bugs during its build sessions, proceeds to fix fake bugs then after the fix is implemented it realize the bug was never real. Claude models waste your tokens
2026-09-04 17:48:16
1
100 :
depends on model & prompts...I use Gpt Sol for the prompts, and depending on the goal...Sonnet is good, but sometimes need to go all way to Fable 5 (but that limits my day goals because of limits) ...if ya buy the hype, ya gonna be disappointed.
2026-09-05 01:57:41
1
FAD :
Started with Lovable, currently using Claude and Codex, but with Claude I mainly use Claude to code rather than using Claude code😆. I know it is anti productive but it’s better for my learning
2026-09-04 15:43:50
1
To see more videos from user @alanonai, please go to the Tikwm
homepage.
