@HypernymAI

We build breakthroughs in long-context for infinite memory + advanced inference solving model scale, context depth & lost in the middle.

 1 AU
hypernym.ai/
Joined August 2024
  • Tweets233
  • Following32
  • Followers51
14Media
Super exciting to see a market starting to be created for AI optimizations with Jev. More to come @HypernymAI
1
6
172
Reading the wrong books can ruin your life, especially if you are open minded. Prompt injection with no sandbox.
40
26
2
381
16,407
Breakthroughs in compression, LITM, hallucination suppression (what we call at @HypernymAI "coherent refusal"), and breakthroughs in recall, when combined, solve this. This curve inverts. The fix already exists, it’s just not evenly distributed - yet.
First Squawk
@FirstSquawkJun 20
COMPANIES CURB AI DEPLOYMENTS AS SOARING COSTS SPARK ROI CONCERNS — FT AI SPENDING SHOCK: FIRMS REIN IN USAGE AS COSTS OUTPACE EXPECTATIONS — FT CORPORATES MOVE TO CONTROL AI COSTS AFTER RAPID ADOPTION WAVE — FT
1
2
9
1,528
Persistent memory has been one of AI’s longest-standing challenges, but practical breakthroughs in context compression and inference are making reliable recall achievable. Persistent memory recall at incredibly fast speeds has already been achieved at Hypernym. I think 4-8 weeks before an actual system that largely solves long-running memory is production ready. Total recall remains the ultimate endgame @HypernymAI
1
2
10
288
Sometimes chaos is the best classroom. You can’t simulate that kind of pressure. You only watch what survives it. 📸 Judging the @wandb Hackathon this weekend, hosted by @sammakesthings and @altryne from the W&B team.
1
2
8
413
If ur in SF for tech week- come join us! luma.com/3gnckox8

Hypernym after hours · Luma

We’re taking all the AWS & Meta buzz and dragging it (willingly) to a secondary location with arcade games, better lighting, and fewer panel discussions. This…

luma.com
1
4
282
LLM Generative AI Frontier Lab Benchmarks; We've been collecting these for a minute. They become more useful every release. I've made it a bit challenging - Which of them do you believe is #Anthropic's #ClaudeOpus41? Which of them do you believe is #OpenAI's #GPTOSS120? Let us know in the comments! This series is the neural reconstruction (read our whitepaper) fidelity test of a 2024 political rally text chunk. We can literally benchmark absolute chaos. Talk to us. calendar.app.google/x8GNRGFx…
7
2
9
283
Replying to @nathan_covey
I think we need to distinguish coding as syntax vs coding as thinking. AI is great for people who think in code but not necessarily keep up with latest language or frameworks. I don’t think you can make useful software without using the prompts as a special type of declarative or descriptive programming language. It is plausible in my opinion that we end up with a new flexible but yet formal programming language to increase the efficiency
42
32
4
818
91,211
Load more