Rob Imbeault

Effort is the algorithm

@TodayInSportsCo This doesn’t exist that any level today

Kids today just don’t understand.

I could pay attention to the road or I can watch a demo on context engineering by @micheltricot @AirbyteHQ FSD-ftw

We Just Shipped What OpenAI, Google, and Anthropic Have Not. Here Are 6 Updates: dev.to/jon_at_backboa…

@ProductHunt middle out

holy sh*t they actually did it

This is what I’m seeing at the rock face. And what our vision is to create an infrastructure that is deterministic in order that AI agents can build on something foundationally strong and trustworthy is our future. We also have some more fun things about our sleeves. Keep building. Keep taking big swings.

I love this! Devs need to chill. "I’m Learning AI in Public, and I Think Developers Need to Chill a Bit" by Jonathan Murray #DEVCommunity dev.to/jon_at_backboa…

@PeriklesGREAT The best

"There is this crazy thing that Generation X did"...🤣

I guess I’m old fashioned. Make the best product that helps people. Market it with integrity. Sell it for a fair price. No stunts. Just a solid platform.

@elliott__potter The perfect amount of nostalgia. Might have aged myself 😝

@RobImbeault thank you Rob! All in-house and lots of late nights...

We’ve raised $27M for this moment: starting today, your agent gets an iPhone and can talk like a friend. Texting is the universal interface. Billions of people text every day, but until now, developers have been restricted from building on the most powerful channel to ever exist. Linq is a single API for iMessage, RCS, SMS, voice, and even FaceTime and Find My. Nothing for users to download. Nothing new to learn. We’re already powering @interaction, @pika_labs, @getlindy, @zocomputer, @joindimension, Tomo (and others we can’t name just yet) to bring this new ecosystem to life. Join them, and start building for free in our sandbox, linked below. Or comment and we’ll get you set up.

@ctjlewis Delve, Cursor, now Supermemory.

I’ve seen enough. No you didn’t. Don’t even need to click.

@manthanguptaa It's cluely level bs. Just wanted engagement.

Love what Dhravya has been building and his work. He understands memory, so I am a little taken aback by this article. I am going to push back on this pretty strongly. This is being framed as a breakthrough when it's mostly benchmark engineering + system inflation. First, 115k+ tokens are no longer a hard problem. We already have 1M+ context windows in production. Calling this "long-term memory" is stretching it. Second, LongMemEval is no longer representative of real systems. It's been around for a while and doesn’t capture long-running agent workflows, real-time updates, noisy tool outputs, and cost/latency constraints. Third, the "99% accuracy" claim is misleading. If you run 8 parallel prompts and count any correct answer as success, you are not improving memory. You are increasing the probability of a hit. That's not intelligence but sampling until something works. (spray and pray) Same with the 12-agent voting setup. It's essentially majority voting across multiple attempts, which inflates benchmark scores without improving core capability. Fourth, saying "no vector DB needed" is not an actual unlock. You have just replaced retrieval with multiple LLM passes, which require higher compute, greater latency, and higher cost. It is shifting complexity, not removing it. These things matter a lot in a production system, so you can forget about deploying this in production. Fifth, the core claim that "agentic retrieval beats vector search" is oversimplified. The real problem in memory systems isn’t embeddings vs agents. Its relevance filtering, temporal consistency, memory lifecycle (what to store/forget), and grounding vs. hallucination. None of these are solved here. Also, this entire system assumes perfect extraction during ingestion, relies heavily on prompt engineering, and offers no guarantees of consistency across runs. So calling memory "solved" is very premature. This is a good experiment in agent orchestration, not a memory solution.

@kimmonismus @grok how does backboard.io compare?

So cool: Supermemory 99% on Sota Memory! •Achieved ~99% on LongMemEval_s using experimental ASMR (Agentic Search and Memory Retrieval) technique. •Replaced vector search and embeddings with parallel observer agents extracting structured knowledge across six vectors from raw multi-session histories. •Deployed specialized search agents for direct facts, related context, and temporal reconstruction; no vector database required. Will be open source in 11 days!

@sarahwooders Agreed. Ps the Letta flat file post was inspired. 🙌

Memory in the sense of recalling information is a solved problem, or at least as solved as it needs to be. That's why everyone is getting ~100% on all the meaningless "memory benchmarks". Memory in the sense of learning/improving over time is very much unsolved though.

@courtlandleer @grok how does backboard.io compare?

it’s hard for me to express sufficiently just how dishonest it is (if you really work on memory) to present a LongMem score as some kind of breakthrough it’s a 3 year old benchmark, the results here are dishonest, it’s a marginal amount of tokens in contemporary ai, an everyone aces it already

@DhravyaShah Hahahahahah what’s funny is that benchmarks don’t go that high so super impressive. Got the engagement you were looking for I guess.

@JayGenXer This is such a no brainer for Canada. Open up your market 10X!!

A FANTASTIC MESSAGE to Mark Carney from USA 🇺🇸 Senator John N. Kennedy

Fucking Legend. 💔

🙌