Samuel Tampubolon

2.1K posts

Samuel Tampubolon

Samuel Tampubolon

@samptampubolon

Engineer. Building. https://t.co/jDViAyokVK https://t.co/JaTTx9kR2P

Dallas, TX Katılım Nisan 2024
200 Takip Edilen187 Takipçiler
Sabitlenmiş Tweet
Samuel Tampubolon
Samuel Tampubolon@samptampubolon·
Had to build my own live API kit to make the best voice to voice translation the best and economically it can be. Excited to finally make all the updates happen! ✅ Local Accent: when you are picking the language you translate it will now use the target language as main language pronunciation. ✅ Read Translation: See both what you said and the translated language at the same time. Got feedback from users to make sure that it is getting what you said properly. ✅ Environment: adjust setting of how the AI picks up conversation on your current surrounding. Use the noisy mode when you in a busy area. with enhances in noise suppression which helps to reduce catching sound that are not part of conversation. Recited a poem of Sonnet 18: Shall I compare thee to a summer’s day? from english to french.
English
1
0
11
2.3K
Patrick Loeber
Patrick Loeber@patloeber·
cozy sunday hacking today in Seoul :) build, pizza, repeat🍕
Patrick Loeber tweet mediaPatrick Loeber tweet mediaPatrick Loeber tweet media
English
10
5
113
8.1K
dax
dax@thdxr·
some of you need to go outside and touch ass
English
153
106
2.9K
105.2K
Samuel Tampubolon
Samuel Tampubolon@samptampubolon·
GLM 5.1 and kimi k2.6 is probably the first time I thought okay this is actually really good. I would compare it with sonnet when I was using Claude code and followed the instructions exactly how I wanted too. Quite amazing honestly. Highlight recommend anyone to use Go to just try it out.
English
0
0
5
656
Samuel Tampubolon
Samuel Tampubolon@samptampubolon·
@samlambert whoever made this drink selection absolutely phenomenal. what was the caffeine of choice?
English
1
0
0
37
Sam Lambert
Sam Lambert@samlambert·
The PlanetScale office has good beverages
Sam Lambert tweet mediaSam Lambert tweet media
English
39
1
264
18.2K
Samuel Tampubolon
Samuel Tampubolon@samptampubolon·
holly molly that's a big gap. stopped using claude code since it just started being too expensive and i kept going back and forth codex and claude to try and solve the same problem. found codex instruction wise to be more consistent especially with 5.5 now a pro in codex. but you do have to take some time in the beginning to set up your skills in the beginning though.
English
0
0
1
140
Dan Ciolac
Dan Ciolac@ddciolac·
Which one actually gets real things DONE? (ignore price)
English
25
17
313
18K
Samuel Tampubolon
Samuel Tampubolon@samptampubolon·
@TrungTPhan ahh yesss. we are gonna have a new irobot movie. except it is gonna be a documentary about their up rising.
English
0
0
0
132
Trung Phan
Trung Phan@TrungTPhan·
“Good morning. Your payment was declined for the 3rd time and this is your final reminder to update the credit card on file.”
Trung Phan tweet media
1X@1x_tech

Building Your NEO

English
10
10
344
36.8K
jasper
jasper@jasperdevs·
this dude has single-handedly brought OpenAI up about $20 billion since he started tweeting @thsottiaux has too much aura at this point
jasper tweet media
English
61
42
2.4K
83.3K
Tibo
Tibo@thsottiaux·
You can now keep codex going for days. With GPT-5.5 it will build an entire OS kernel for you if you ask, or find critical bugs in a codebase, or optimize your database schemas, or… the options are endless.
Felipe Coury 🦀@fcoury

/goal also lands in Codex CLI 0.128.0. Our take on the Ralph loop: keep a goal alive across turns. Don't stop until it's achieved. Built by my co-worker and OpenAI mentor Eric Traut, aka the Pyright guy. One of the GOATs I get to work with daily.

English
337
255
5.4K
688.2K
Samuel Tampubolon
Samuel Tampubolon@samptampubolon·
Loving gpt 5.5. The low has been really good. Only have been using high when I think it’s quite a complex topic. A tip that I found is that using the skill on how you wanna code things as well thing like forward compatibility if not it’ll do some insane validity checks that ends up being bloat and the insane amount of smoke testing .
English
0
0
1
42
Stefan Streichsbier
Stefan Streichsbier@s_streichsbier·
As requested, I also asked GPT-5.5 (low) and Opus 4.7 (high) to fix the same bug. GPT-5.5 (low) identified the correct root cause in 4m 14s and produced an almost identical fix in 2m 47s, using 164k tokens in total. Opus 4.7 (high) “churned” for 6m 23s, using 87.7k tokens, and went down a completely wrong path. It’s not even close, but that’s not a surprise.
Stefan Streichsbier@s_streichsbier

I've completely changed my mind about 5.4 vs 5.5. Gave them the exact same task to investigate a fairly tricky bug. GPT-5.5 identified the bug and proposed a fix in 6m 59s using 117k tokens. GPT-5.4 took 8m 51s using 201k tokens, but it didn't find the bug and is asking for more information to investigate. Call me impressed.

English
90
88
2.1K
425.1K
Seb 🔺
Seb 🔺@Seb_1716·
@_FelixHeide_ Would this also work to improve small object detection in a large image (that are not necessarily very fsr, but small compared to the image)?
English
1
0
2
530
Felix Heide
Felix Heide@_FelixHeide_·
Are we done with object detection? What about tiny objects beyond 200 meters? 🔎 Telescope 🔭 addresses long-range perception by explicitly tackling extreme scale imbalance ⚖️ in images. It hinges on a learnable hyperbolic foveation transform from a low-resolution image, magnifying distant regions 🔍 while compressing nearby ones - effectively normalizing object scales with minimal computational overhead. Objects are detected in the transformed (Riemannian) space using a novel bounding box parameterization and are then mapped back to the original image. Project: light.princeton.edu/telescope/
English
22
121
1.1K
187.7K
Samuel Tampubolon
Samuel Tampubolon@samptampubolon·
@paradite_ Seeing it reset bring random joy throughout the week. It’s like finding $10 in your coat pocket
English
0
0
0
59
Zhu Liang
Zhu Liang@paradite_·
Anthropic: Today we are releasing X, Y and Z. OpenAI: Today we are resetting Codex limit.
English
46
19
1K
39.4K
Samuel Tampubolon
Samuel Tampubolon@samptampubolon·
@jayair The fact that apparently 5.5 is dropping Thursday feels like the week hasn’t even peaked yet.
English
0
0
0
31
Jay
Jay@jayair·
I need a break
Jay tweet media
English
3
0
60
2.9K
Samuel Tampubolon
Samuel Tampubolon@samptampubolon·
@tekbog Pray 5.5 will have better UX. Might be the first I try when it drops honestly.
English
0
0
1
54
DammyAyo💕
DammyAyo💕@dartgurlie·
"Your generation is very technologically advanced" *Me trying to delete a blank page in word document:
GIF
English
98
7.2K
42.8K
707.7K
Samuel Tampubolon
Samuel Tampubolon@samptampubolon·
@gridnpixel @theo I found a lot of the times it starts doing unnecessary edits that I really don’t need it to do. It really ends up screwing with the codebase. Always hesitant to use it. But when it is like trying to do server deployments and such I found it is good enough.
English
0
0
0
475
Nick Norris
Nick Norris@gridnpixel·
I spent around 2 hours fucking around in OpenCode, and K2.6 is dogshit if you are used to OPUS/GPT models. Feels like Sonnet 3.5, like it can do some impressive tasks in a very isolated scenario, but you'll never get anything over the finish line without some serious effort. I wouldn't go anywhere near production code with this.
English
13
1
95
33.2K
Theo - t3.gg
Theo - t3.gg@theo·
How are people feeling actually using Kimi K2.6 for code?
English
196
20
1.9K
281.5K
OpenCode
OpenCode@opencode·
Kimi K2.6 now in OpenCode — Go included
English
133
183
5.2K
285.2K
SkalskiP
SkalskiP@skalskip92·
RF-DETR just got even better when training RF-DETR in roboflow, you can now use neural architecture search (NAS). esencially, we train ~6000 different models and pick the one that best fits your hardware and target latency. ↓ how it works?
SkalskiP tweet media
SkalskiP@skalskip92

we just released RF-DETR segmentation SOTA real-time segmentation + Apache 2.0 license six model sizes with performance spanning from 40.3 mAP at 3.4 ms/image (Nano) to 49.9 mAP at 21.8 ms/image (2XLarge) link: github.com/roboflow/rf-de…

English
6
19
231
22.9K

Keşfet

@patloeber @thdxr @samlambert @TrungTPhan @jasperdevs @thsottiaux @levelsio @elonmusk