AI geek turning ideas into reality. Always ready for the next ride. @Nickš
San Francisco
Joined August 2022
- Tweets1.1K
- Following170
- Followers7
- Likes6.1K
Nick Davis retweeted
I love that AI has made it possible to do completely ridiculous and unscalable things like build + maintain a custom icon library
Nick Davis retweeted
My TLDraw agent is getting better
Pretty arrow-heavy, but not bad. Nice use of colour in particular
Nick Davis retweeted
39 years after Black Monday, Wall Street got a bounce instead of a crash
On Oct 19, 1987, the Dow plunged 22.6% in a single session, still the worst day in US stock market history
This Monday, the weekend chatter was full of crash talk after last week's AI selloff. Into the open, Nasdaq 100 futures were up about 1.2% and S&P 500 futures about 0.5%, per Reuters
Last week was brutal. The Nasdaq fell 4.5% and the S&P 1.95%, the worst week for both since late June. The Philadelphia Semiconductor Index dropped for 8 straight sessions, its longest losing streak since October 2008, closing Friday at 11,137 after a 3.8% drop and down 16% since Oct 6
$CRWV, which had plunged more than 25% last week on debt fears, climbed about 6.6% in premarket. $AVGO and $MU were up more than 3% each, with $NVDA and $AMZN up more than 2%
Oil cooled too. Brent was down about 1.5% near $108 after some Russia sanctions were waived, taking a little pressure off rate-sensitive tech
The real test is still ahead. $NFLX and $SPCX report Tuesday, then $META and $TSLA on Wednesday. Last week's selling was about how the AI buildout gets financed. One quiet open on the anniversary of 1987 does not answer that. It just shows that when everyone is positioned for a crash, the bounce often shows up first
ALT Dark chart comparing last week's Nasdaq down 4.5%, S&P down 1.95% and SOX down 11.4% with Monday Nasdaq futures up 1.24% and S&P futures up 0.45%, next to the Philadelphia Semiconductor Index's 8-session losing streak matching October 2008, noting SOX down 16% since Oct 6 and the 39th anniversary of Black Monday 1987 when the Dow fell 22.6%
Nick Davis retweeted
November 3rd ā Forge '26 ā San Francisco.
This is probably one of the biggest AI events in the world.
NVIDIA's Jensen Huang and DHH are hosting fireside chats. @FireworksAI_HQ's CEO, Lin Qiao, is doing the keynote.
The event is free, but you need to apply to attend.
Impossible to summarize everything you'll get out of this day, but in short, this is the time and place to learn how the best teams in the world choose their models, design evals, serve them, and route requests.
You'll learn about what worked, what failed, and how those decisions affected quality, latency, and cost.
To send an application and get a free ticket, use this link:
fandf.co/4yugvYz
Thanks to the Fireworks team for partnering with me on this post.
Nick Davis retweeted
> saturday: new chinese model
> monday: another new chinese model
> both almost frontier level
> both cheaper than lunch
the AI price war just turned into a massacre
MiniMax M3.1 Flash: Space Performance + Bunny Speed
šš° Yes, Space Bunny - which has topped global leaderboards with 200T+ tokens - was an earlier version of M3.1 Flash.
š°Redefining the performance economics of agentic AI. M3.1 Flash achieves frontier-level intelligence across key tasks such as Cowork (Automation-Bench: 62%; ALE: 30%), Coding (DeepSWE: 74%) and Visual Reasoning (BabyVision: 91%), at 1/3 the cost of M3 and 3x the speed.
API pricing (flat rate): $0.15 / $0.60 per 1M input/output tokens, with cache reads at $0.015 / 1M tokens.
Model weights will arrive on Oct 23.
API: platform.minimax.io
M Plan: minimax.io/m-plan
MiniMax Code for desktop: agent.minimax.io/download
Nick Davis retweeted
In 2027, developers who care more about ābeautiful codeā than user experience arenāt gonna make it
Nick Davis retweeted
This reminds me of Yazi/DuckDB glory.
If you want to read S3 with Yazi (and a DuckDB extension), you can just browse and preview the files as if they were local. v2.0 makes this even better š.
Check out the YT video at: youtu.be/cimUvBd_9Ns
Or read the write-up: ssp.sh/brain/s3fs-fuse/
Tech is so amazing, and so is DuckDB!
ALT Simply browse S3 Buckets locally with Linux File Browser (yazi) mounted via s3fs-fuse
DuckDBās upcoming 2.0 release looks to make reading files from S3 comparable to just reading a database.
Even with Lakehouses, reading directly from object storage is a performance bottleneck. Itās not fast.
Thereās normally a caching layer involved. But Iām seeing DuckDB 2.0 changing all that.
Nick Davis retweeted
build housing or face your city being overrun with LinkedIn claudeslop poasters, retatrutide dropshippers, and cultists of the machine god [real]
Nick Davis retweeted
The slowest version of the new @Tesla Roadster (<1.6s to 0-60 mph) will be quicker than the current quickest version of the all-electric $2.5 million Rimac Nevera R (1.66s).
The fastest trim of the Tesla Roadster (rocket trim) will be capable of 0-60 mph in ~0.59 seconds.
Nick Davis retweeted
If you can build real life wall hacks
You can build real life single frame aimbot
Much to think about
you don't even have to be in the same room
with the right Wi-Fi setup, someone walking behind brick walls can still leave a measurable trace at the receiver
the graph uses real Wallhack1.8k recordings: orange is walking, blue is an empty space, both on the same scale. the 3D scene illustrates the idea
if your next thought is āi'll just sit still,ā watch the quoted postš
Nick Davis retweeted
AI backlash is coming.
So many people we see think they are AI gods aren't using it well at all, just wasting huge amounts of money without producing much value.
The ROI won't be there for many companies.
They'll clamp down on token spend.
Everyone will get mad.
Nick Davis retweeted
TypeScript is the new TypeScript š
When ššššššš-compiled, outperforms tsgo by 2-4x
Nick Davis retweeted
Cindy is one of the best distributed systems engineers I know, and resistant to any and all hype (also a great writer!)
From her:
"Claude (when given proper instructions, constraints, and context) now writes better code than me, almost always."
Nick Davis retweeted
"A decent model with a great harness consistently beats a great model with a bad harness."
@addyosmani wrote that, and 857K people saw it on X.
Addy spent over a decade on Chrome at Google and now works on Claude Code. When someone who builds these tools says the setup matters more than the model, I listen.
So how do you actually build a great harness? Here's the roadmap I'd follow if I were starting today:
1. Learn what a harness is
Agent = Model + Harness. It's everything that isn't the model.
2. Write good instructions
The file your agent reads every single session
3. Give it real context
Your product, your users, your data
4. Build skills
Playbooks it loads only when it needs them
5. Add memory
So you never correct it twice
6. Set permissions
What it can do alone, and what it never can
7. Connect tools
Your files, your terminal, your MCP servers
8. Add checks
Grade the agent before a customer does
9. Close the loop
So it keeps going until the work passes
10. Put it together
Start with your own work, where mistakes are cheap
The roadmap in the image has a real case study for every part, from OpenAI to Ramp.
Every product is getting an agent. This is how you build one that works.
ALT How to Build AI Agents That Actually Work: the harness engineering roadmap, 8 parts with a case study for each.
Nick Davis retweeted
US oil rigs fell by 3 to 459 this week per Baker Hughes, even with Brent near $105.60 and WTI at $97.56. The total rig count held at 603, while EIA data showed crude output dropping to 13.679M bpd from 13.979M. #OilPrice #RigCount #Oil #Shale
oilprice.com/Energy/Energy-Gā¦
Nick Davis retweeted
Iāve been trying to get better at spotting talent before itās obvious.
Thereās no better person to learn from than @apartovi
He founded @neo and backed Facebook, Dropbox, Airbnb, Cursor, and Kalshi.
When Cursor looked like a worse version of GitHub Copilot, the founders couldnāt convince some of their own friends to join. They knew the team was brilliant. Some were already using the product. They still said no.
Ali spent hours trying to change their minds. He didnāt know exactly how Cursor would beat GitHub either. He believed the team would figure it out.
What was he seeing that everyone else was missing?
On this week's episode of The Library of Minds, we talked about how he developed that judgment, the warning signs you ignore when hiring, and who is liable when AI breaks the law.
What I like about Ali is how openly he talks about getting things wrong too: blowing a negotiation with Steve Jobs, backing down after threats from Jimmy Iovine, and what he wishes heād handled differently in his public feud with YC.
1:35 - How to Judge Talent in 10 Minutes
7:50 - His #1 Red Flag in Founders
9:05 - How to Spot Top Talent Early
14:30 - Should Kids Still Learn to Code?
22:05 - Nobody Wanted to Join Cursor
31:10 - He Tried to Trick Steve Jobs
40:30 - Extorted for Half His Company
47:45 - The Y Combinator Feud
53:10 - When AI Commits a Crime
63:25 - Choose the Path You Fear
Nick Davis retweeted
Updated the chatbot-template to use the new Trace component. If you want to see how to use it with @aisdk how to wire questionnaire etc, check the repo.
github.com/shadcn-ui/chatbotā¦
I just published Trace, a new component for agent activity. It's one abstraction for reasoning, tool calls, queues, chain of thought, subagents...etc
Also in this release: Code Block and AI Elements.
ui.shadcn.com/docs/changelogā¦
Nick Davis retweeted
You don't need 10,000 followers to get AI clients.
One of my students landed a $5k/month retainer with about 50.
Here's the content system that makes clients come to you š§µ
Nick Davis retweeted
Iāve never seen Tesla people hype anything like theyāre hyping the Roadster launch today
Nick Davis retweeted
Thank you for this wonderful conversation, Dara! š¤©
Iāve been trying to get better at spotting talent before itās obvious.
Thereās no better person to learn from than @apartovi
He founded @neo and backed Facebook, Dropbox, Airbnb, Cursor, and Kalshi.
When Cursor looked like a worse version of GitHub Copilot, the founders couldnāt convince some of their own friends to join. They knew the team was brilliant. Some were already using the product. They still said no.
Ali spent hours trying to change their minds. He didnāt know exactly how Cursor would beat GitHub either. He believed the team would figure it out.
What was he seeing that everyone else was missing?
On this week's episode of The Library of Minds, we talked about how he developed that judgment, the warning signs you ignore when hiring, and who is liable when AI breaks the law.
What I like about Ali is how openly he talks about getting things wrong too: blowing a negotiation with Steve Jobs, backing down after threats from Jimmy Iovine, and what he wishes heād handled differently in his public feud with YC.
1:35 - How to Judge Talent in 10 Minutes
7:50 - His #1 Red Flag in Founders
9:05 - How to Spot Top Talent Early
14:30 - Should Kids Still Learn to Code?
22:05 - Nobody Wanted to Join Cursor
31:10 - He Tried to Trick Steve Jobs
40:30 - Extorted for Half His Company
47:45 - The Y Combinator Feud
53:10 - When AI Commits a Crime
63:25 - Choose the Path You Fear