Deep Tech Solutions. Hardware. Software. #AI, classical, #quantum, Designer of Fidelity Framework (formerly SpeakEZ.Tech). #cleflang
United States
- Tweets 2,464
- Following 297
- Followers 166
- Likes 5,102
Braidpoint retweeted
Its crazy how much of the discourse around agents is just astrology for SWEs
12
36
591
10,364
RT @GaryMarcus: boom, there it is.
or perhaps i should say, “bust, there it is”
all this insanity about demand and there are literally w…
111
1
Braidpoint retweeted
POV: Channeling my inner Rihanna because I won my case against the @washingtonpost — and they owe me a year’s worth of back pay. 👊🏾💰
298
1,255
11,111
187,262
🦆 We made a @duckdb tool (using 2.0 compiled with PGQ) for a fast graph retrieval of 24 repos for Fidelity framework. As code pushes surface the corpus is auto-updated and agent calls into the service always receives the latest cross-indexed info! 🧡 clef-lang.com/blog/retrievin…
1
16
Braidpoint retweeted
I wonder if ad-supported business models will need a revisit since anyone can just strip them out of products now
todays target is a baby camera
asked DeepSeek to root it, de-bloat it, and cut it off from the cloud. it verified the whole thing across two reboots
finally I can delete the app thats filled with ads
1
2
204
Braidpoint retweeted
This is freakin' crazy - so DeepSeek has 10 month payback on GPU purchase, despite being 20 times cheaper than OpenAI/Anthropic. And they are enabling everyone + every company to have your own AI, your own inference, your own data.
Cancel the fucking IPOs 😂 This is so insane 😂
Wow 🤩 The whale 🐳 isn’t just shipping highly efficient, affordable frontier models - they’re open-sourcing the production inference stack that actually serves them at scale.
DeepInfer (Apache 2.0) powers DeepSeek-V4.1-Flash w the same high-th/put, low-latency runtime used in production. It includes optimized kernels, Attention DP / balanced CP, MoE EP + EPLB, DSpark speculative decoding w confidence-scheduled verification, SWA bounded replay (to cut prefill work), 0-overhead asynchronous scheduling, batch-invariant inference, deterministic sampling, + structured outputs.
Offline inference examples run from local Hugging Face checkpoints w no network required once the weights are cached. They also open-sourced deepseek-recipe, the production tokenizer/encoding path so community deployments match official API behavior.
Open models + open kernels + open production inference. That’s a meaningfully different trajectory the whale 🐳 is offering to the world. I freakin love these guys 💙🥹💙
29
115
1,397
54,879
"obsessive" == doing the basics consistently
When our AI code reviewer's cost per run doubled, we dug into the token usage to find out why. Learn how we used prompt caching and intelligent model routing to slash our inference bill by 68% while speeding up reviews. cfl.re/4isj2NU
19
The expectation (and perhaps one that's naive) is that a late-breaking Y2K style realization will filter out across the industry, and when senior managers can't blame-shift then *something* will start moving forward. Whether that's 'enough' at today's velocity (relative to 1997) is perhaps the question we're all *really* asking by proxy. braidpoint.tech/blog/the-mos…
1
15
