Philip Tompkins

Husband, Father, Founder - Builder of things.

[Philip Tompkins @Philbert440]

I'm really enjoying watching this play out 🍿

[Artist_Icecream 🥁 @Artist_Icecream]

This is my CEO

[Philip Tompkins @Philbert440]

@elonmusk bravo

[Philip Tompkins @Philbert440]

🤣

[Nikita Bier @nikitabier]

@vibhu There are a few user groups that misused Following as a way to promote their accounts (See: Crypto) with no intent to actually engage with or read the account’s content.

Their Timelines are now destroyed beyond all recognition.

[Philip Tompkins @Philbert440]

Agreed, barrier to entry is lower than ever, but the biggest challenge of building a business hasn't changed and is still just as difficult: finding product market fit.

[@levelsio @levelsio]

All you really need today to build a business is 100% free open source software that charges you $0/mo

+ a VPS server
+ an API to do some required AI stuff
+ some R2/S3 file hosting

They don't want me to tweet that because it destroys their businesses but I can't lie to you!

[Philip Tompkins @Philbert440]

@Comedyorwat Lol, thanks for the heads up.

[Philip Tompkins @Philbert440]

Can we come up with a new word for AI "not slop" where a creative human driving AI makes something genuinely inspiring or useful?

@SamyazaSpeaks I'm curious, how long did this take you? This is very cool.

[Samyaza Speaks @SamyazaSpeaks]

The Biblical Experience

Music: Holy Place by BENNETT

[Philip Tompkins @Philbert440]

@Comedyorwat The difference between gamblers and investors get more of the latter and less of the former into your algorithm

[Philip Tompkins @Philbert440]

@Scobleizer I personally will, but not everyone has a server rack and space for it to be noisy away from people and power setup for it

[Philip Tompkins @Philbert440]

Maybe the counter point it we all become homelabers running models on old datacenter hardware as it gets cycled out, 8x 32gb nvlink'd v100s can run some pretty decent models and there is an emerging community like 1Cat-vLLM that's maintaining good inference support for the cards, when A100s get cheap enough that'd be 8x 80gb nvlink'd so the large models become viable.

[Philip Tompkins @Philbert440]

@felixrieseberg https://t.co/USBMyMOPoV

[Philip Tompkins @Philbert440]

@theo I've had decent luck with android but I have tried much on iOS yet

[Philip Tompkins @Philbert440]

I made a visualizer for my https://t.co/USBMyMOPoV agent nodes last night. I just had the idea and threw together a prototype to see if I could, I like it. Still some cleanup work to do but I'll add this to the public repo soon.

[Philip Tompkins @Philbert440]

@xai @SpaceX with all this Fable drama going on, I can't help wondering when are we going to get @grok 4.5 with cursor data @elonmusk

[Philip Tompkins @Philbert440]

@ThePrimeagen @jamwt About $70k and a proper 240v circuit
https://t.co/lVHzfPi3fE

[Philip Tompkins @Philbert440]

@ThePrimeagen I did the same a while back, still love it

[Philip Tompkins @Philbert440]

@gregisenberg 1. my own custom harness (hermes less bloated)
2. grok+cursor data (anthropic)
3. grok-build (I really don't like either but claude-code)
4. local models❤️but cloud is more capable.
5. bootstrap till scaling then VC speed up market capture.
6. best time
7. job growth

[Philip Tompkins @Philbert440]

@Chuks89Einstein @APompliano @cfosilvia Not really vertical AI though it's still a general cloud model under the hood, it proves that a harness, MCP servers, and context can got you better model output on specific related tasks.

[Philip Tompkins @Philbert440]

@theo correction: grok-build+composer

[Philip Tompkins @Philbert440]

@theo that green check you are here now is exactly where i am with my usage except replace open ai with local models, grok + composer is surprisingly good and I'm looking forward to a grok model trained on that data