yikes (:D/acc)
@yikesawjeez
Followers
3K
Following
5K
Media
359
Statuses
11K
sees sharks, jumps. a generalist agent. agent chop shop, erry wed https://t.co/0aK1UfdZ4v @zos_sfm, @latentspacepod @tribecodeai fmr: @swellnetworkio, @elizaos
staring @ github trending
Joined January 2021
i have a memecoin but i'm not telling you the contract address
1
0
3
so i thought what i'd do was i'd just pretend i was one of those deaf-mutes i'd just have people write everything on a note and shove it over to me
0
0
0
Personal Computers emerged from (3) stacked-innovations → microprocessors (smaller) → operating systems (useful) → visicalc (killer app) Personal AI will emerge from; → smaller GPU → fast inference engine (new OS) → killer app (?) We spent two years, solving all 3
25
5
218
boys r gross & girls just want to lay eggs in your chest cavity or whatever, back to relative emotive language weights in localstorage till the haptic feedback is finished for me, thanks
There are a ton of women in sf, they just don’t want to hang out with you
1
0
1
i am very tired of these words 'prompt' 'vibe coding' 'mcp' (most of the time) 'agent'(ic, but slightly less so, unless incl. 'gradient', then ok) 'tool' 'sub-agent' more, but i digress your mental model is backwards they can be so much more than lil robot ppl if u let em
0
0
0
┏┓ ┃┃╱╲In this ┃╱╱╲╲ house ╱╱╭╮╲╲ ▔▏┗┛▕▔ ╱▔▔▔▔▔▔▔▔▔▔╲ We trust open-source LLMs ╱╱┏┳┓╭╮┏┳┓ ╲╲ ▔▏┗┻┛┃┃┗┻┛▕▔
9
3
70
the mental model is your part of the job, meat
vibe coding with Solid.js: - syntax ~= React. most models are very familiar and do well - because the syntax is seemingly similar to React, it fail at the mental model wrong (e.g. signals != state, components are not at runtime, components run once, jsx is an element)
1
0
1
As a fun Saturday vibe code project and following up on this tweet earlier, I hacked up an **llm-council** web app. It looks exactly like ChatGPT except each user query is 1) dispatched to multiple models on your council using OpenRouter, e.g. currently: "openai/gpt-5.1",
I’m starting to get into a habit of reading everything (blogs, articles, book chapters,…) with LLMs. Usually pass 1 is manual, then pass 2 “explain/summarize”, pass 3 Q&A. I usually end up with a better/deeper understanding than if I moved on. Growing to among top use cases. On
772
1K
15K
> i leaned in and told him > listen, kid... the only AI you can trust is the one you run yourself > you don’t own your compute? > then you’re out here kissing somebody else’s ring > he looks at me confused, > so i spell it out > nobody hands you compute in this life > otherwise
11
9
64
sf is an unfair advantage & it should be as accessible to estonian gas station attendees w a used thinkpad on their smoke break as it is the full-ride berklee dropout at frontier tower
- Zero network - Zero capital - Zero warm intros - Zero US connections - Zero badges: no YC, a16z, Stanford, MIT - Zero safety net This was our superpower. Immigrants who made money by bootstrapping mobile apps, willing to sacrifice everything and work 100x harder than anyone
0
0
0
the least valuable thing you have your software engineers doing is writing code
Coding is the easy part, your grandma could do it Marketing is the hard one
0
0
0
> sees new RAG is dead blog post and opens it (shame on me) > see the word knowledge graph > promptly close the post
12
2
131
i made a @peterthiel bot to rip the pitch deck a new one id link it but the algo will tank me so ig engagement farm, comment or rt or smth if u want a link heres sum of its feedback, lil guy came up with a good few heaters
0
0
0
The 3 stuck founder archetypes: 1. The vertical ai believer raised 3m to build AI for dentists or contractors or whatever spent 18 months learning the industry they shouldve known before starting built something that works but customers pay slow and churn fast now at 400k
22
9
200
i love ahmad greentext threads
> you are > a normal dev wondering how to become an Applied AI Engineer > you heard “embeddings” and “RAG” 1000x > want to know what they actually are, how they plug into LLMs > first: what even is an “embedding”? > embedding = a list of numbers (a vector) that represents
0
0
1
you and claude are smarter together than either one of you on your own
AI is getting so good that what I recommend for 2026 is that people start including training as part of their agent system. So a significant task is not complete until your AI system has explained how it works and used Socratic questioning to make sure you understand. We can
0
0
1
d&d minis, competitive EDH, selling m: tg proxies to ppl at the card shop bc i hate wotc & wanna play cEDH, also the 4th thing but i forget what it is now gibe check, link in bio
If a founder spent their childhood: – Grinding ranked in any competitive game – Building random shit with a 3d printer or Raspberry Pi – Selling stuff online (anything from Pokemon cards to dropshipping) -Making Minecraft or roblox games Invest in them immediately.
0
0
1
very honored to briefly bring @thdxr out of paternity leave for an incredible online track keynote from one of the most transparent, brutally honest builders in our space! taken in one shot, no less. the casual flex...
@zRdianjiao @Zai_org Our second Online keynote is live from @thdxr of @OpenCode, arguing that the real hard problems of building AI products have not changed: https://t.co/ttyXaJGpp5 While AI can generate endless slop, it cannot generate cool and "aha" moments. This requires deep taste, not
9
10
318
stop waiting to get saved and start saving yourself lets go out lets get goin
In January, my startup had 13 days of runway left when I realized no one was coming to save me VC who’d agreed to bridge us pulled out the day wire was due Cut burn 80%, signed 1st annual contract w 1 day left before payroll & somehow got profitable Hardest thing I ever did
0
0
3
ngl i stan graphite hard but also use my product lol
Code Reviews won't ever be the same. AI can check your code 100x faster and catch 10x more issues than anybody can. There are two remaining reasons for a manual code review: 1. To transfer knowledge within a team 2. To ensure AI didn't miss critical aspects in the code Go to
0
0
1