Rafe

755 posts

Rafe banner
Rafe

Rafe

@RafeAphd

A.I. Reseracher. Prev- @join_ef || @OfficialUoM || @UniOfYork || @imperialcollege In a past life I pursued kick-boxing. 6"4. 220 lbs. i try i try

London شامل ہوئے Mayıs 2024
330 فالونگ79 فالوورز
پن کیا گیا ٹویٹ
Rafe
Rafe@RafeAphd·
Solo dev. Starting this challenge from scratch. Goal: 100k MRR in 12 months building micro-SaaS products with AI. No team. No VC. Just shipping and seeing what sticks. First product is live now. Will be scraped in 1 month. Posting everything, the wins, the flops, the numbers.
English
2
0
6
374
Rafe
Rafe@RafeAphd·
@shl end state is forward deployed customer support
English
0
0
0
195
Sahil Lavingia
Engineering is eating customer support
English
9
2
77
10.8K
Rafe
Rafe@RafeAphd·
@anthdm this is common knowledge and i think is papers too
English
0
0
0
21
Anthony GG
Anthony GG@anthdm·
GLM 5.2 is 100% distilled from Opus. Change my mind.
English
107
6
534
79.9K
Rafe
Rafe@RafeAphd·
@Teknium did you just do the sakana thing
English
1
0
1
449
Teknium 🪽
Teknium 🪽@Teknium·
Introducing Mixture of Agents 2.0 in Hermes Agent. Combine any provider's models into a mixture of your own. Access your presets as if it were a normal model in Hermes. Big improvement in our soon-to-release HermesBench against opus and gpt-5.5 with MoA using Opus & GPT together.
Nous Research@NousResearch

The strongest models are gated and access is granted only to a select few. Hermes Agent now exposes MoA presets as virtual models, giving you capabilities beyond the publicly available frontier: 8% higher than Opus 4.8 and 11% higher than GPT 5.5 on our upcoming benchmark.

English
96
117
1.4K
208.6K
Rafe
Rafe@RafeAphd·
@scaling01 I keep getting PTSD to crypto tokens
English
0
0
0
99
Rafe
Rafe@RafeAphd·
@louis030195 in wrapping of satire exists tiny pockets of truth
English
0
0
0
9
Rafe
Rafe@RafeAphd·
gpt 5.6 will run at 750 toks through cerebras if you can get access
Rafe tweet media
English
0
0
1
40
BridgeMind
BridgeMind@bridgemindai·
GPT 5.6 Sol will run on Cerebras at 750 tokens per second.
BridgeMind tweet media
English
12
8
385
18.1K
Rafe
Rafe@RafeAphd·
@zerohedge will hardware get commoditised that quickly are you sure about that ?
English
0
0
0
36
zerohedge
zerohedge@zerohedge·
Can we fast forward 2 years when everyone will have a local model running on a 10TB DDR8 rig (which will cost $29.95)
English
162
207
4.8K
265.4K
Rafe
Rafe@RafeAphd·
@beffjezos yeah will have to wait so long for the release tho
English
0
0
0
11
Rafe
Rafe@RafeAphd·
@mxfp4 congrats dude you climbing the ranks!
English
0
0
0
16
Kevin Wang
Kevin Wang@mxfp4·
aaaand its flagged on hackernews, anyone delt with this before?
Kevin Wang tweet media
English
1
0
3
132
Rafe
Rafe@RafeAphd·
@JiweiLi1 amazing work, testing now will be back with a review
English
0
0
0
69
Jiwei Li
Jiwei Li@JiweiLi1·
Excited to share Ornith, our latest family of open-source models specialized for agentic coding. Ornith achieves SOTA performance among open-source models of comparable size on a variety of coding benchmarks (Terminal-Bench 2.1, SWE, NL2Repo, OpenClaw, SWE Atlas, etc) Feedback is deeply appreciated! 📖Tech Blog: deep-reinforce.com/ornith_1_0.html 🤗Huggingface: huggingface.co/collections/de…
Ornith@ornith_

Aloha! 🌺 Meet Ornith-1.0, a family of open-source LLMs specialized for agentic coding. Ornith-1.0 spans the full parameter sizes including 9B Dense, 31B Dense, 35B MoE, and 397B MoE. It achieves state-of-the-art performance among open-source models of comparable size on coding benchmarks including: ✅Terminal-Bench 2.1(77.5) ✅SWE-Bench(82.4 on verified, 62.2 on pro, 78.9 on Multilingual) ✅NL2Repo(48.2) ✅SWE Atlas(41.2 on QnA, 42.6 RF, 39.1 TW) ✅ClawEval(77.1) Post-trained on top of gemma4 and qwen3.5, Ornith-1.0 employs a novel self-improving training strategy in which reinforcement learning is used to generate not only solution rollouts, but also the task-specific scaffolds that drive those rollouts. By jointly optimizing the scaffold and the resulting solution, the model generate higher-quality solutions in agentic coding.😎 All models are released under the MIT license, enabling full commercial and research use. 📖Tech Blog: deep-reinforce.com/ornith_1_0.html 🤗Huggingface: huggingface.co/collections/de…

English
50
39
538
41.6K
Rafe
Rafe@RafeAphd·
@vineerpasam buying up everyone favourite apps, if its isnt broken dont fix it
English
1
0
1
23
Vineer
Vineer@vineerpasam·
Where is he now?
Vineer tweet media
English
9
1
12
349
Rafe
Rafe@RafeAphd·
@SStricklandMMA sean this is serious character devleopment i would have never seen coming
English
1
0
2
147
Sean Strickland
Sean Strickland@SStricklandMMA·
Everyday I wake up and want a drink or drugs Some days I feel so disconnected, bored and empty and the thought of a beer or a little weed just to ground me, make me present is so justifiable in my mind.... But I know where that road leads. Just say no and endure.
English
1.7K
1.2K
28.6K
1.2M
Rafe
Rafe@RafeAphd·
@thoughtcrime___ i think pricing any lower currently wouldn't line up with unit economics
English
0
0
0
463
Rafe
Rafe@RafeAphd·
@mxfp4 its made for a very specific audience
English
1
0
0
41
Kevin Wang
Kevin Wang@mxfp4·
reminder that the hackernews login page looks like this
Kevin Wang tweet media
English
9
1
16
915
Rafe
Rafe@RafeAphd·
@juhapellotsalo thanks dude! still working on improving the harness
English
1
0
0
8
Rafe
Rafe@RafeAphd·
Solo dev. Starting this challenge from scratch. Goal: 100k MRR in 12 months building micro-SaaS products with AI. No team. No VC. Just shipping and seeing what sticks. First product is live now. Will be scraped in 1 month. Posting everything, the wins, the flops, the numbers.
English
2
0
6
374
Rafe
Rafe@RafeAphd·
@chaosengineerr yes for dopamine reasons i cant get into haha
English
1
0
1
38
Wahab Khan
Wahab Khan@chaosengineerr·
let's be honest, would you still build if it paid nothing for a year?
English
26
1
15
1.3K
Rafe
Rafe@RafeAphd·
@saloniiio its all review and evals
English
0
0
0
14
Saloni
Saloni@saloniiio·
In 5 years, what will be more valuable? Writing code or Reviewing AI-generated code.
English
72
7
52
3.2K
Rafe
Rafe@RafeAphd·
currently thinking of hiring our first engineer, is it too early
English
1
0
2
55