skeptrune @skeptrune Twitter profile

Last Seen Profiles

@ClintDinnel

@itirafet0

@UKHouseofLords

@dayotetede

@FirstGentleman

@AH_Scriptum

@tonygurney

@Coach_MCollins

@TomHart99697994

@j_i_hamilton

@FAZ_Feuilleton

@chmpgnsoshalist

@avneetdhillon

@becca2uk

@NicoleEinbinder

@x3LUCKY3

@eamonnryan41

@FMGSuite

@LindseySmithAZ

@_kendimce_7

@GreatAlvierCR

@GerkenSascha

@booksss_da

@saint1jcee

@iammillestate_

@MdHasan77963800

@Josh_Lane1

@epongandam

@LCSBaseball

@wisbechrugby

@hinari_egg

@MillCreek_MBB

@Sheng_Peng

@UOCuniversidad

@_xa1f

@FoxcatcherBet

skeptrune

@skeptrune

3 months

Another good reason for why Rag is here to stay

45

61

459

skeptrune

@skeptrune

23 days

@baileysimrell MX Master 3 is actually very nice

3

0

93

skeptrune

@skeptrune

2 months

Sub 10ms semantic search anyone? 🤘 Yesterday we realized that the 30ms required to make the embedding was coming from initializing a reqwest http client in Rust and not inference. Once we switched to ureq, it dopped to 3ms. 🔥 Major thanks to @jobergum !

5

73

skeptrune

@skeptrune

3 months

This repo by @kelseyhightower just made my day a lot brighter 😂😂😂

3

6

70

skeptrune

@skeptrune

4 months

@lcamtuf @BrendanEich such a good demo of where we actually are with AI 😂😂

1

0

59

skeptrune

@skeptrune

2 months

Sub 100ms semantic search anyone? It has arrived!

6

1

49

skeptrune

@skeptrune

2 months

POV: you are about to get 10x better search

3

35

skeptrune

@skeptrune

2 months

We hit #2 HN front-page and I was asleep! 🥹

3

1

32

skeptrune

@skeptrune

4 months

@thdxr semantic search > RAG is the entire thesis of our company

1

0

26

skeptrune

@skeptrune

2 months

We have finally done the YC Launch! Is the upvote button working for you guys? 👀👀🤔

2

1

26

skeptrune

@skeptrune

1 month

Grooving

Jake Duth

@JakeDuth

1 month

Haven't looked at this in a while, but holy crap I've never seen it this green on my own profile

1

0

9

4

1

25

skeptrune

@skeptrune

4 months

we got the beanbag it's official, we're a tech company 😂

2

1

23

skeptrune

@skeptrune

1 month

High quality demo day production value 🤣🤣

4

0

24

skeptrune

@skeptrune

3 months

Trieve on the right and Algolia on the left, early preview before I sleep 😴😴😴 still relevance tuning and UI bugfixes to do, but it's looking good so far 😁🚀

1

2

22

skeptrune

@skeptrune

2 months

Easily crushing 100 RPS on ingest rn and search times are completely unaffected 🧙😎 3 months ago each of these requests took 500ms.... 👻

1

0

22

skeptrune

@skeptrune

2 months

We are about to make $10k MRR look like peanuts 🏃‍♂️ There's no longer a latency disadvantage to picking Trieve (left) now 🚀

4

1

20

skeptrune

@skeptrune

2 months

Woooooohoooo! Don't forget to also star the github 👀🌟

Y Combinator

@ycombinator

2 months

. @TrieveAI (YC W24) is an all-in-one infrastructure for search operations. Engineers implement a pipeline connecting their data source to Trieve's backend for non-technical staff to fine-tune relevance with no code. Congrats on the launch, @skeptrune &…

2

4

45

9

1

20

skeptrune

@skeptrune

2 months

This is the best UI I have seen built on @trieveai by FAR!!! Check these guys out!

Y Combinator

@ycombinator

2 months

Glimmer (YC W24) is a new way to search massive PDFs using AI. If you've ever dealt with huge PDFs, you know how broken search can be. Glimmer lets you search PDFs in natural language and get answers instantly. Congrats @GillBates408 + @prantheman__ on…

7

93

1

2

19

skeptrune

@skeptrune

2 months

500!!!!!!!!!!!!!! Let's goooooo! 🚀🔥

0

3

16

skeptrune

@skeptrune

2 months

Sometimes semantic search really is magical 🧙 Imo, the internet is going to be a much better place once every search bar works like this

1

15

skeptrune

@skeptrune

3 months

@dannymoerkerke damn this is depressing i thought that I couldn't possibly hate Apple to any greater extent, but it turns out that my resentment can grow much more

0

15

skeptrune

@skeptrune

2 months

I have no idea what caused this growth, anyone else know? Woke up to 100+ new stars

2

1

15

skeptrune

@skeptrune

2 months

Building RAG systems in the past year, chain-of-thought and other techniques always seemed non-viable because I don't think users will tolerate 5s+ latency in the optimal case. Groq's inference time is going to make a lot of these applications workable I think.

Groq Inc

@GroqInc

2 months

"With AI, rapid inference can make the difference between halting interactions and real-time spontaneity... Fast LLM inference can be immensely useful for building agents that can work on problems at length before reaching a conclusion." -- @AndrewYNg

4

2

29

5

2

15

skeptrune

@skeptrune

2 months

ngmi if you're not arguing about GrapheneOS and writing customer on-boarding docs at 11pm 😎🤔

Liz Wessel

@lizwessel

2 months

Why has no Search API company come close to dethroning the main incumbents? They all were started around 2012 (or earlier), and it seems like developers often complain about them. (alternatively, LMK if Im just not thinking of a company who has indeed taken off.)

4

0

10

5

0

14

skeptrune

@skeptrune

1 month

I'm going to be talking about building search and RAG for an OpenAPI spec Thursday at 8AM PT with @qdrant_engine Schedule a notification at the link below!

1

15

skeptrune

@skeptrune

3 months

@GroqInc is amazing! Will be switching Trieve to it ASAP. Idc how much better Lllama2/GPT5, etc. is. Latency is king 👑. I'd rather help users iterate on a prompt quickly over and over again than marginally improve ability to get it right the first time w/ better LLM.

0

1

13

skeptrune

@skeptrune

4 months

@thdxr used pgvector?

1

0

14

skeptrune

@skeptrune

6 months

@Jupiluxe @SeekTheFinds At the point where you are having interactions which even remotely resemble this something has gone horribly wrong. I think I'd be in an asylum.

0

13

skeptrune

@skeptrune

4 months

a community member made a Medium post about getting started with our current rather clutzy self-hosted service ahead of our 0.3.0 release with a managed SaaS offering stuff like is is incredibly cool for me ❤️, i think it makes me happier than sales

3

0

12

skeptrune

@skeptrune

2 months

🚀🚀🚀

Pete Koomen

@koomen

2 months

Nicholas ( @skeptrune ) and Denzell ( @denzell_ford ) want to power every RAG application and search bar on the internet. After graduating college in 2.5 years, these two started out working on a different idea and built their own search infrastructure from scratch after trying…

2

29

3

0

12

skeptrune

@skeptrune

2 months

Look under the hood of any great company and you shall find Kirkland 😂 @greptileai

1

0

13

skeptrune

@skeptrune

2 months

The Y-Combinator team is missing out on applicants because Algolia can't tell you they're hiring when you query for it. We are making it way easier for companies to take advantage of the technological leap in information retrieval on display here.

0

2

12

skeptrune

@skeptrune

1 month

@andersonbcdefg @twofifteenam > slaps the machine on a gurney That's AGI right there

0

13

skeptrune

@skeptrune

1 month

Of all things for NYT to diss, it's PostgreSQL?!?!

2

0

12

skeptrune

@skeptrune

3 months

search with Algolia on the YC Companies directory isn't great "cloud storage" does not return "Dropbox" "application error monitoring" does not return "PagerDuty" made a little script to grab all the company URLs so I can build a new version w/ Trieve 👇

2

0

10

skeptrune

@skeptrune

7 months

@briannekimmel Who/where are these investors? We would like to talk. $30k ARR at 100% month/month growth since launch w/ patent-pending moat and still struggling to close our round

0

skeptrune

@skeptrune

5 months

@peter I am most definitely the squirrel

1

0

12

skeptrune

@skeptrune

4 months

i like our new branding

Trieve (YC W24)

@trieveai

4 months

Welcome to Trieve We are currently in the works of rebranding Arguflow! Get ready for some changes this week;)

2

1

6

2

0

11

skeptrune

@skeptrune

11 days

Stuffed the beanbag in the backseat to go to the new office 😎😎 Startup type beat 😂

2

0

12

skeptrune

@skeptrune

2 months

Makes me sad that more people don't see this as a failure of CS programs

varepsilon

@var_epsilon

2 months

crossing the student -> hacker bridge can either be done by spending 5 years at bigco or building a bunch of side projects

195

281

7K

6

0

11

skeptrune

@skeptrune

5 months

@Jupiluxe @Tyson_James_ Lol, you committed a crime by putting this in my feed w/ a reply

1

0

9

skeptrune

@skeptrune

2 months

@qtnx_ It's always been this way and always will be in every field imho

0

11

skeptrune

@skeptrune

3 months

My hope for generative multimedia AI is that it gets easier to produce things like this and we end up with a lot more of them 🙂

Ariel Helwani

@arielhelwani

3 months

Damn. The promo for AJ x Ngannou is a literal movie.

245

2K

16K

1

2

10

skeptrune

@skeptrune

3 months

I do believe retrieval is valuable and am excited about what I am trying to do with Trieve! However, I do want to clarify that I am much more bullish on retrieval than I am augmented generation on the whole. Big R little ag!

Robert Lukoszko — e/acc

@Karmedge

3 months

And… BIAS

1

26

1

0

11

skeptrune

@skeptrune

2 months

We are going to replace Algolia so fast. Hoooly How tf is this acceptable from a vendor you're paying > $5k/mo

1

10

skeptrune

@skeptrune

9 months

How in the hell did @wagieeacc do @LegionOfSkanks w/out me knowing about it?!?!

2

10

skeptrune

@skeptrune

6 months

@tekbog We call this tilt

0

11

skeptrune

@skeptrune

4 months

huddled around the devops wizard deploying onto the GPU server, very nice cheese = good

1

0

10

skeptrune

@skeptrune

2 months

Broadly I'm also just a lot less bullish on LLMs. Embedding models, sparse encoders, and cross encoders are AMAZING, but LLMs are not as good The useful thing here are the search results. I wouldn't even bother waiting for the LLM to load /shrug

Owen Colegrove

@ocolegro

2 months

I have been obsessed with LLMs since ChatGPT came out. No experience since first getting cable internet has compared. I've experimented a lot with what to build and landed on a framework, R2R, to help devs build better RAG systems. Here's why I think it's so worthwhile -

3

4

96

2

1

10

skeptrune

@skeptrune

25 days

Just use Trieve If information retrieval is important to you then it's likely worth picking a source available vendor that has your back and gives you full control. Github below!

Simon Willison

@simonw

26 days

I'm begging you @OpenAIDevs , please tell us how your chunking works! It's a small detail, but it makes a huge difference in helping me make decisions about how to effectively use your RAG implement

15

17

326

1

2

10

skeptrune

@skeptrune

8 months

@OpenSourceOrg @HeatherMeeker4 @OSSCapital Kind of. More important to create quantitative metrics that can quickly evaluate the quality of RAG'ed inferences and notify the user of hallucinations.

0

3

skeptrune

@skeptrune

3 months

first comparison blog just dropped!

Trieve vs. Algolia

Compared to Algolia, Trieve offers a more powerful source available search platform with dense-vector semantic capabilities, RAG support, and a ranking plus relevance fine-tuning dashboard for...

trieve.ai

1

0

8

skeptrune

@skeptrune

3 months

Semantic search is really fun sometimes! Should probably remove the companies tagged as inactive from our dataset, but not terrible results for this query tbh.

2

1

9

skeptrune

@skeptrune

26 days

@eatonphil Every other Friday at NelNet we got to hack on whatever we wanted and record a demo video. It was a nice fun day where you could pick up a new skill or fix some DX issue that bugged you. Sharing it with the rest of the team after was always the best part.

3

0

8

skeptrune

@skeptrune

2 months

I felt a lot more comfortable shamelessly asking for upvotes when we were bootstrapped 😂

0

7

skeptrune

@skeptrune

3 months

damn! finally got a span going for our search route and now we know why it's slow 🤯 turns out we optimized the hell out of our @qdrant_engine integration long ago and the real issue is with creating the embedding for the search query

2

0

9

skeptrune

@skeptrune

1 month

Approve and merge

Siddhant Khare

@Siddhant_K_code

1 month

What would you do to PRs like these as an OSS project maintainer?

177

4

248

0

9

skeptrune

@skeptrune

14 days

One of the nuanced things that mentally stuck from YC was @pedroh96 's talk mentioning that Brex started as a CLI. If Trieve's dashboard UI started as a CLI instead, we would have launched faster with a higher quality API. No UI burden + closer to API dogfood'ing.

1

0

8

skeptrune

@skeptrune

2 months

Hard agree

fks

@FredKSchott

2 months

The trend in infra of "rug-pull your free plan after you've gained adoption" is slimy and needs to stop

18

21

369

2

9

skeptrune

@skeptrune

3 months

I am much more excited about cites than "putting 100M tokens in the context window." Much of time users will likely want to read the retrieved chunks as well as the generated inference if the inference is good. If 100M tokens are in the context window that is impossible to do.

1

8

skeptrune

@skeptrune

2 months

Would anyone else like to see a Trieve demo in the form of a search and RAG engine for Philosophize This by @iamstephenwest ?

0

2

7

skeptrune

@skeptrune

1 month

Going live!

skeptrune

@skeptrune

1 month

I'm going to be talking about building search and RAG for an OpenAPI spec Thursday at 8AM PT with @qdrant_engine Schedule a notification at the link below!

1

15

2

1

9

skeptrune

@skeptrune

2 months

Fun stuff, Trieve in the infra category 🚀

1

0

9

skeptrune

@skeptrune

3 months

Honestly we might just have to stop offering OpenAI as an option for creating the dense embedding vectors They consistently spike latency like this and it really damages customer experience for us

4

0

8

skeptrune

@skeptrune

2 months

@ycombinator @usearini @abduljamjoom @rami_rustom Bullish on focusing a voice assistant into a single vertical. Imo, it makes it a lot easier to create a good product

0

8

skeptrune

@skeptrune

3 months

Search is an important problem, Algolia couldn't find this...

Yuri Sagalov

@yuris

3 months

It feels like all the ingredients are there to create a completely AI-powered podcast. The first/easiest one to create would be something like NPR's Up First podcast — A 10 minute summary of the three biggest stories of the day, tailored to your interests. Anyone building this?

10

1

21

2

6

skeptrune

@skeptrune

4 months

LFG

2

0

8

skeptrune

@skeptrune

2 months

This is a fairly good architecture diagram for Trieve. We just add a few more services for performance and scalability.

Gwen (Chen) Shapira

@gwenshap

2 months

It has come to my attention that some of you have not realized how central the database is to RAG implementations. So I drew it out for y'all. As Hubert said: "It is a database + LLM"

3

25

165

0

8

skeptrune

@skeptrune

5 months

@ThePrimeagen but who gets the credit when they succeed?

2

0

8

skeptrune

@skeptrune

15 days

This is likely already an improvement over a coding assessment for screening a candidate. Plus odds are it can get a lot better with a few improvements like down weighting the massive .ipynb files for languages pie chart and better recognizing tech stacks. Exciting! 🚀🚀

GitRoll

@gitroll_io

15 days

#GitRoll 's @ycombinator W24 Top 10 Notable YC Tech Founders #8 Nicholas Khami: 8.18 Mid-level Full-stack Developer @skeptrune founder-engineer at @trieveai ranked in the top 10% Get your GitRoll score now:

0

1

7

1

8

skeptrune

@skeptrune

4 months

@patricksrail @emollick @TwoWeeksataTime i think that is exactly what the op is trying to avoid here lol that sounds terrible

0

8

skeptrune

@skeptrune

23 days

🚀🚀🚀

Scott Condron

@_ScottCondron

23 days

@EnriqueGuerraF @cerebral_valley Thanks for the recommendation but on my way to the airport now. Ended up meeting with @skeptrune and learning about what theyre building at @trieveai . Looking forward to using it as a search component within a project, looks slick

2

1

4

1

0

8

skeptrune

@skeptrune

3 months

> Founder-engineer I think I'm settled on using this term to describe my role at Trieve (the company I'm currently working on). Seems to fit nicely and I feel good about it.

6

0

8

skeptrune

@skeptrune

4 months

@WarrenInTheBuff does this mean more Rust jobs? it should mean more Rust jobs

1

0

8

skeptrune

@skeptrune

3 months

I want to build a browser extension with one of these super long context models that tells me where the most interesting information is.

1

0

6

skeptrune

@skeptrune

3 months

This is what I mean when I say that the dense vector embedding models go so much harder than the LLMs You can totally do this by embedding sentences and comparing the resulting vectors looking for max, min, outliers by space, etc. Big R little ag!

Visakan Veerasamy

@visakanv

3 months

once you have a mid-sized body of work, lets say ~200,000 words, you can start getting some interesting additional value out of rereading old material looking for phrases that have a surprising energy to them. i think this is something that will be difficult for chatgpt to do

8

4

172

1

6

skeptrune

@skeptrune

2 months

I'm a big believer in this for most things that are early. Obviously not all that viable at scale, however, delay as long as you can imo

2

0

7

skeptrune

@skeptrune

4 months

when talking to people that likely know what Algolia does, i think i'm just going to start saying "we're an open source alternative to Algolia" it's easy and clean and let's everyone know what we're doing seems to work

3

0

6

skeptrune

@skeptrune

2 months

Very fluff post imo. Are they tuning cross-encoders, LLM's, embedding models? Not sure. /shrug maybe it'll be good, maybe not

Contextual AI

@ContextualAI

2 months

Today, we’re excited to announce RAG 2.0, our end-to-end system for developing production-grade AI. Using RAG 2.0, we’ve created Contextual Language Models (CLMs), which achieve state-of-the-art performance on a variety of industry benchmarks. CLMs outperform strong RAG…

35

140

1K

2

0

7

skeptrune

@skeptrune

29 days

This is the best part of using Rust when you don't even have validation that what you're building is useful and are whimsically tossing code to main. It provides a lot of confidence in shipping fast without doing much review or planning.

Eric Fredine

@fredine

29 days

@o__boga And this is why C needs to be used with caution. The first Rust solution forces you to consider the out of range case. It’s also possible the code emitted by the Rust compiler is as efficient as the jump table - especially with assertions.

1

0

6

1

0

7

skeptrune

@skeptrune

9 months

@lxsmnsyc @FUTO_Tech You do seriously great work man!

0

7

skeptrune

@skeptrune

25 days

Wow, this sounds like everything I have ever dreamed of

3

0

7

skeptrune

@skeptrune

15 days

Trying to learn a bit of awk for fun! 🚀🚀 Just went from: `docker ps -q | xargs docker kill` To: `docker ps | awk '{print $1}' | tail -n +2 | xargs docker kill`

1

7

skeptrune

@skeptrune

3 months

@kelseyhightower There's a release! 🤣🤣💀💀

0

7

skeptrune

@skeptrune

25 days

@n_s_bradford @OpenRouterAI Might start a new thread where I do this more consistently. Also, OpenRouter is a seriously amazing product. Worth using just for the nice unbloated UI.

1

2

7

skeptrune

@skeptrune

2 months

The OG hat 😎😎😎 There are still ~6 remaining 👀👀

blee¢h

@drnk_bleech

2 months

who ya think you talkin to, a turkey?!?🦃 from @goodcharls stream last night 🥂

7

8

93

1

0

6

skeptrune

@skeptrune

6 months

@tekbog Carmack who?

1

0

7

skeptrune

@skeptrune

3 months

I'm very interested in how the PDF function for Gemini is chunking the text into messages. For it to be this good, I would expect the approach is more evolved than just shoving all the text into the context window completely unstructured.

4

0

6

skeptrune

@skeptrune

2 months

@ycombinator @trieveai Major thanks to @FUTO_Tech for helping us get started! ❤️❤️

1

7

skeptrune

@skeptrune

2 months

Wow, this is the most organized we have been on project planning in a long time Feelsgoodman

0

7

skeptrune

@skeptrune

3 months

just had our first ever well thought out question in Discord, very exciting 😎😁😁 i love open source!!!!!!!!!

1

0

7

skeptrune

@skeptrune

3 months

Just realized I haven't shared this. The most magical semantic search demo we have done to date is search for collegiate debate evidence. I give you "fish are friends not food" ✨✨✨

0

6

skeptrune

@skeptrune

4 months

Remix going the opposite way of Next lol

Remix 💿

@remix_run

4 months

Remix is not just for server-rendered apps anymore! Introducing Remix SPA Mode, new in Remix v2.5

23

149

887

2

0

6

skeptrune

@skeptrune

3 months

1 year of hard grinding with our code out in the open! 💪🚀

Trieve (YC W24)

@trieveai

3 months

Three Hundred Stars Thank y'all!

1

0

4

1

6

skeptrune

@skeptrune

2 months

Real Algolia copy says 6 months to get value!?! Damn the bar is low

0

4

skeptrune

@skeptrune

2 months

Get it up end to end on Trieve with sub 50ms retrieval performance in less than 10mins. We support both intermediary search query and no intermediary out of the box and you can adjust the prompt for it on a per-dataset level it in the dashboard UI.

Owen Colegrove

@ocolegro

2 months

RAG is not that hard to deploy, but we should aim to make it trivial. Next, we should aim to make it easier to iterate and scale. The best LLM app builders I have spoken w/ are still doing the basics. Everyone *wants* to do the sexy new thing, but in reality no one has time.

3

54

1

0

6

skeptrune

@skeptrune

3 months

Trieve on the left, Algolia on the right 😎😉 DM me if you would be game to beta test it and give us feedback!

1

0

5

skeptrune

@skeptrune

2 months

@mattturck Seems like a solid ad. What do you feel is so uniquely good about this?

1

0

5

skeptrune

@skeptrune

3 months

was hunting for reference Launch YC posts and found this gem @brian_armstrong (I think) looking for a co-founder in 2012 and getting mildly flamed in the comments very inspiring and cool