Kai
@hqmank
AI Builder ๐ ๏ธ
Practical AI Workflows & Coding Tips
High-Value Promo Codes Updated Regularly
DM for collab ๐ฉ
245 Following 11.6K Followers
SuperGrok Heavy no longer comes with free Cursor Ultra.
You only get Grok Bot usage now.
Cursor Origin is live. Their own code hosting, built for AI agents.
This won't displace GitHub. Open source community, Issues, PRs, Releases, package ecosystem, all do real work Origin does not touch. Origin handles one thing: multiple agents committing and merging in parallel at scale.
An extension of Cursor's ecosystem. Relevant if you already run Cursor agents daily.
Show more
Origin, our code hosting platform, is now live.
It's fast, easy to use, and deeply integrated with Cursor.
Get started by syncing your repos from GitHub.
Which AI subscription is the best value right now?
ChatGPT Pro: $200/month
Claude Max: $200/month
SuperGrok Heavy: $300/month
If you work with AI, or hire people who do, read "The AI Engineering Skills Map."
Andrew Ng identifies four core skills:
1 Building and deploying AI apps
2 Software engineering fundamentals
3 Using coding agents
4 Shaping what gets built
Required reading for the AI era.
Show more
New: A map of the most important skills in AI Engineering.
DeepSeek raised its API prices today.
At peak hours, V4 Pro cached input went from $0.0036 to $0.044 per million tokens. Thatโs 12x.
V4 Flash cached input is up 5x. Uncached input is 3x and output is 4.5x across both models.
The cheap V4 preview era ended fast.
Show more
DeepSeek raised prices.
OpenCode Go cut its $15 DeepSeek V4 Flash quota by 70%.
DeepSeek has officially increased their prices
OpenCode Go's limits have been updated to reflect them
We have made progress on operation cheepseek, more info soon
GPT-5.6 Sol's 1M context in Codex now works with ChatGPT accounts.
Leave it off for small tasks. Default is already tuned for speed and cost.
~/.codex/config.toml
model = "gpt-5.6-sol"
model_context_window = 1000000
model_auto_compact_token_limit = 900000
Show more
GPT-5.6 Sol 1M in Codex. This used to only work for API keys, but we just flipped the switch and works for usage through ChatGPT accounts now too.
The same warning applies, there is a reason the current context length is the default, we have tuned it to ~perfection. But you do you!
Show more
Just tested it.
Doesn't work anymore. Skip it.
SuperGrok Heavy is $99/month for 6 months, then $300. Looks worth it to me.
You get Heavy and X Premium+. Link Grok in Cursor and you also get Ultra ($200) and Grok Bot.
New signups only.
Link ๐
Show more
SuperGrok Heavy is $99/month for 6 months, then $300. Looks worth it to me.
You get Heavy and X Premium+. Link Grok in Cursor and you also get Ultra ($200) and Grok Bot.
New signups only.
Link ๐
Show more
GPT Astra is about to drop this week.
Codex might reset.
Codex
โ
Almost 100% reliable
โ
Occasional resets
โ
Open-source
โ
(will have Astra)
Quick clarification: the $80 reset fee applies to ChatGPT Pro accounts.
That makes more sense. At $80, you could get a few Plus accounts instead.
ChatGPT Pro users will soon have to pay $80 to reset their weekly Codex quota.
Would you pay?
๐จSoon, resetting your weekly Codex limit will cost $80.
Would you pay?
OpenAI has something big on the way.
Will a Codex reset come with it? ๐
Incredible things are happening at OpenAI right now. Energy is high.
SuperGrok changed its Cursor offer.
SuperGrok Heavy users used to get one free month of Cursor Ultra. Eligible users can now keep Ultra free for as long as their Heavy subscription stays active.
If Heavy gets Ultra, regular SuperGrok should come with Cursor Pro too @elonmusk
Show more
Unsloth just turned local AI into a desktop app.
Run or fine-tune models on your own machine, then connect them to Claude Code or Codex through an OpenAI-compatible API.
Kimi K3's open weights are dropping in a few hours.
It could be the first open-weight model that can truly compete with the best closed models.
A 3T-class model built for coding, agents, and reasoning.
I'm excited to see what the open-source community does with it.
Show more
DeepSeek uploaded DSpark checkpoints for DeepSeek-V4 Flash and Pro.
Same checkpoints, with a speculative decoding module attached.
DeepSeek claims 60%-85% faster generation for Flash in live traffic.
"Spark" is doing exactly what the name says.
Show more
