Alp (@49agents) header
Alp (@49agents) profile picture

Alp

@49agents

Building Qorqbot @ 49Agents

49agents.com/qorqbot
Joined February 2026
  • Tweets22.3K
  • Following23
  • Followers898
4.8KMedia
Photo by Alp (@49agents)Photo by Alp (@49agents)Photo by Alp (@49agents)Photo by Alp (@49agents)Photo by Alp (@49agents)Photo by Alp (@49agents)Photo by Alp (@49agents)Photo by Alp (@49agents)Photo by Alp (@49agents)Photo by Alp (@49agents)
Financial data is RIPE for JEPAs... noise, non-stationarity, unknown ground truth states... yet data and evaluation is not readily available! We resolve that in our latest NeurIPS paper `Towards Financial World Modeling' TLDR: invest with LeJEPA ;) 📄arxiv.org/abs/2610.09048 🧵⬇️
Image from Randall Balestriero (@randall_balestr)
2
32
4
237
25,612
today we're launching DEALS 🎁✨ 630 deals on tools you will actually use worth over $3.3 million combined → Price: FREE → buildinpublic.com/deals what deal would you like us to add next?
Image from Build in Public (@buildinpublic)
135
33
6
500
147,206
Claude Opus 5.5 just took a big drop on NerfBench. Yesterday it was scoring above launch. Today it's at 94.2%. GPT 6 Astra: 98.0% Sonnet 5.5: 100.9% GPT 6.1 Sol: 106.7% 94.2% is still inside normal variance, so we can't call it a nerf yet. But we're watching Opus 5.5 very closely.
384
341
166
6,849
942,709
Alp retweeted
Artem (@sl1ma4) profile picture
Artem
@sl1ma4Sep 6
POV: you assigned 4 Gemini agents to fix a single bug.
184
491
54
12,300
577,619
Top 5 launches today on StartupBase 1. 49Agents IDE (@49agents) 2. LongTerMemory AI Study Tool 3. MaturityHR 4. Kekoso (@a_chmerev) 5. Fix.Pictures (@farri_io) startupbase.io/launches/dail…

fix.pictures AI — Amazon Product Photo Fixer

Turn any supplier photo into an Amazon-ready listing image. AI removes backgrounds, corrects framing, and exports 2000px compliant JPGs in seconds.

fix.pictures
2
1
2
97
"dude Ox Alpha on OpenCode is insane" "yeah 1M context with unlimited usage" "have you tried it?" "no. you?" "no."
Image from pdawg (@prathamgrv)
67
78
12
2,633
82,853
Affirmation of stupidity ruined the entire tech industry
Hannah Rosenberg⚡ (@hmichellerose) profile pictureHannah Rosenberg⚡
@hmichelleroseAug 21
Image from Hannah Rosenberg⚡ (@hmichellerose)
19
54
1
1,122
13,717
Alp retweeted
Alp (@49agents) profile picture
Alp
@49agentsAug 18
Replying to @absencelul @github
honestly the github hate is overblown. they shipped copilot before anyone else and got hit by the agentic wave from both sides - cursor from the left, openai from the right. hard to move fast when you are the target
1
4
53
Alp retweeted
Klaas (@forgebitz) profile picture
Klaas
@forgebitzAug 17
what if cursor made all our agents push to github all at once taking down github launching their competitor
38
3
2
625
20,304
Alp retweeted
Alp (@49agents) profile picture
Alp
@49agentsAug 17
Replying to @Equateall
trydemoday.com
DemoDay — a 20-second film of your product.

DemoDay — a 20-second film of your product.

Drop in your website; we build the film from it. Up to 45s, 1920×1080, loop-clean.

trydemoday.com
1
27
Replying to @49agents @MiaAI_lab
Yes. My rule of thumb: Use reasoning effort low for 90+% of tasks, and xhigh for planning/complex code (or better yet, a stronger model). No reason to use medium reasoning effort. And turning off reasoning completely crates the results, except for very easy tasks.
1
2
113
Alp retweeted
Ankit (@ankitkr0) profile picture
Ankit
@ankitkr0Aug 15
you are 1 pivot away from being a billionaire
a16z (@a16z) profile picturea16z
@a16zAug 14
Cursor, 4 years ago
Image from a16z (@a16z)
46
257
11
4,992
486,101
Alp retweeted
Alp (@49agents) profile picture
Alp
@49agentsAug 12
This is insane... It took me 2 hours to build this production grade demo with Fable. Made it using my own plugin for content creation... If you want the link to the plugin, please leave a comment and I will DM it to you!!
2
1
8
10,262
My code doesn’t have bugs It has unexpected features that only appear in production
14
4
34
1,476
Agent work gets messy fast when every project lives in another terminal tab. 49 Agents IDE is a 2D agentic IDE for managing AI agents, terminals, files, git context, and Beads issues across projects and machines. It helps you keep multi-agent work visible by putting terminal sessions, editor panes, machine status, and navigation onto one zoomable canvas instead of another stack of tabs. Key features: • Zoomable canvas – place panes spatially, resize them, and keep persistent layouts • Real terminal sessions – uses tmux via ttyd with ANSI color, scrollback, and your shell config • Multi-machine workspace – connect agents from different machines to one canvas without SSH • Self-hosted access – run locally, over LAN/Tailscale, or through a relay from laptop, tablet, or phone • Builder tooling in the canvas – Monaco editor, git graph, Beads issue tables, markdown notes, and permission notifications Public GitHub repo under BSL 1.1: free for individuals and small teams, converts to MIT on 2030-02-26. Link in the reply 👇
Image from Dan Kornas (@DanKornas)
4
8
35
2,431
Alp retweeted
Yoan (@yoanwaidev) profile picture
Yoan
@yoanwaidevAug 11
Replying to @49agents @UmarSaeedX @omkarships
yeah this is the real cost, not the model comparison. one screen with all of them and a live status per session, plus answering a blocked one without switching into it. that is what made running two of them cheap for me.
1
2
37
gpt-5.6-sol is overrated
Image from Broooooklyn (@Brooooook_lyn)
35
2
3
248
53,098
Alp retweeted
Alp (@49agents) profile picture
Alp
@49agentsJul 12
I just got Fable (code) spitting out a chinese character. In my experience, cheaper chinese models would have this problem randomly at times. Smells fishy...
Image from Alp (@49agents)
4
1
6
459
Alp retweeted
zek (@zekramu) profile picture
zek
@zekramuAug 8
if you ever feel bad, just remember there are people out there using opencode lol
81
14
8
647
80,960
you, an imbecile, learning to prompt LLMs better me, also an imbecile, learning to prompt google better
Image from illiquidity providooooor (@skyquake_1)
117
1,168
103
33,825
1,691,360
Load more