David Tippett

David Tippett

@taidesu.bsky.social

Open Source enthusiast and builder of search systems. Currently building search at Firecrawl.

352 followers109 following509 posts

Hot take for spooky season: "AI Search" for most companies means "we added embeddings." Embeddings don't learn from clicks. They don't test ranking changes. They don't get better next week. Same label. Different machine. πŸŽƒ

So we asked Jev if people should use Relevan for their Agentic search and the answer was a resounding yes! Then we tried to do some real work with and and were super happy with the outcome. Check out our blog (which links our open source eval repo) for the methodology: relevan.dev/blog/decisio...

Relevan is officially live! The beta is capped so you may get added to the wait list as you login. We are slowly letting people onto the platform. If you're interested in becoming a design partner we still have a few slots open so reach out! Join at relevan.dev

Each company has different data, language, users and goals. Most platforms give everyone the same tools. Relevan is building a relevance loop per customer: measure quality, learn signals, test improvements and tune to their use case. Beta tomorrow: relevan.dev

Agent-first search isn’t a chatbot bolted onto an old console. Relevan uses LLM-driven onboarding. Bring any agent, point it to our LLMs.txt, OpenAPI spec, and per-index skills, then let it configure, ingest, and search. Beta opens Wednesday: relevan.dev

Relevan launches its beta Wednesday, August 19. We’re building search for a world where agents configure it, query it, and help improve it. No more need for a dedicated search engineer. Join the waitlist: relevan.dev (click Join Beta) #AgenticSearch #AISearch

I have come to a newfound hatred of slacks account system. I hated it before but now that Google allowed me to change my email handle slack seems to be the only place that is freaking out. Despite claiming to support OAuth slack requires me to set a password to change the email on my account...

I hate poking the GitHub commits search cluster. That thing is a behemoth and any mess up means probably a month or more of cleanup jobs. That said it needs a lot of work since it's gotten so large. We're doing some much needed segment merge work.

Dang got my first Do Not Contact for my startup. Gotta have pretty tick skin for these types of things. I believe theres a world for better search! Just need to find the people who are aligned with the mission. :D

Prospecting for my startup is eating my soul alive. I got pretty thick skin but not hearing back after all the outreach I have done is killing me. For every 1 outreach I do I have like 7 that I disqualify. I mean we are going for a niche but I didn't realize ti was that niche.

This is definitely the best photo of me from Haystack. For context, I am keeled over as a pain response to someone simplifying my career to flipping on a switch πŸ˜†

Eric Pugh@dep4b.bsky.social

A three part story!

A watched website never breaks. Literally had my marketing site crash twice yesterday for two different people. Loading it up to debug today? NUFING sat there for 15 min like a dummie waiting for it to break.

I just got cup-jacked at Starbucks. The barista accidentally put some one else's drink into my cup and handed it out. Instead of realizing: hey... this isn't a normal Starbucks cup... and it doesn't have a lid... they person just took it. πŸ€ͺ

The barista offered to buy me another cup and I was like no way I am not going to make you do that. A shift manager ended up expensing one for me tho which was nice. Free cup :D

Why do editors, devrel, and the like feel the need to make so many asinine edits to my blogs? Like you changed "With search being" -> "Given search is" Is it more clear? Idk It just rubs me wrong when I've gone through these changes 3 times with each subsequent editor changing the voice.

I am 2 months into building my saltwater aquarium tracker tanki.dev and sent a newsletter out asking for feedback. And my users showed up! #1 bit of feedback? "Make auth sessions longer" - people who track aquariums typically do so only once per week and auth logged out after 7 days.

tanki.devTanki

Along with that another contender for tank tracking software has hit the market. I'm honestly not checking them out at all. I believe in what I am building. I can only hope that they're listening to their users and doing the same. It is encouraging though since it means others see the need.

To be clear I am not just doing customer driven development. My goal is to show them a future they didn't even realize was possible. One where each persons learnings helps create a more whole picture.

βœ… Change checklist ready βœ… Fail back plan ready βœ… Pager assigned to me βœ… Monitoring up ❌ Getting up the gut to make a hard change to avoid an inevitably worse incident tomorrow...

I am doing my best to resist building an operator for Elasticsearch. I dont think I can for much longer though. We need a stable way to automatically change cluster settings. Rolling upgrades? Reduce merge pressure and increase shard move speed. Low QPS times? Lets bump up merge concurrency.

David Tippett reposted

The second edition of Designing Data-Intensive Applications, by myself and @chris.blue, is finished and sent off to the printers! Ebooks should be available in the next week, and print books in 3–4 weeks. Sigh of relief. πŸ˜… (BTW, this is a good opportunity to support your favourite local bookshop!)

Dang. My Compute blade project is literally pandoras box. I initially got a bunch of the Uptime Compute blades to run an OpenSearch cluster. After months of waiting I get the blades in, go to start up OpenSearch and BOOM. Craps out because the Amazon Linux containers can't run on ARMv7. 🧡

View full thread

🧡 CM5's are ARMv8 so a few of them and I can do it! Of course none were available because of the initial shortage. Finally get the CM5s and then I have to wait for the availability of heat sinks. Wait for a while and finally get a community made version. Now to install the CM5's FINALLY πŸš€

🧡 ANNDDDDDDD MY SSD's WONT MOUNT because the Kingston NV2 wont train on the CM5 and buying new drives is not ideal because storage cost is up 3-4x because of the DRAM shortages. Thank you for coming to my Ted talk. 🫑 See you in 10 years when I finally finish this project.

Dang. My Compute blade project is literally pandoras box. I initially got a bunch of the Uptime Compute blades to run an OpenSearch cluster. After months of waiting I get the blades in, go to start up OpenSearch and BOOM. Craps out because the Amazon Linux containers can't run on ARMv7. 🧡

🧡 No worries, others are posting OpenSearch containers right? Get it running for a bit using the bitnami/opensearch containers and BOOM - Bitnami pulls their containers and moves to a paid model -_- (I know I could make my own but I don't want to maintain that infra 😭)

New favorite move. Introduce drama for no reason. Example: I got my wife ice cream. Before handing it to her I assured her no blood was spilled getting her ice cream. The concern that followed was priceless.

I just realized my understanding of the Elasticsearch/OpenSearch refresh + translog operations was INCORRECT. I'd always had the mental model that refresh or translog flush would write segments. That's partially true...

The right way to think about it is to say that Refresh writes segments to disk. Translog "stamps" them as being official (Lucene commit). If the Translog reaches it's flush size it'll trigger a refresh before committing the segment. They're related but not as an either or but a both and

The amount of logistical things needed to build a startup is so frustrating. I don’t want to build a company I want to ship code that helps people. Today I: - created a bank account - migrated emails - mailed so compliance documents