Proximal

@ProximalHQ

Advancing coding intelligence.

4 394Followers

0Following

113Posts total

340.5KViews on collected posts

Against accounts of the same size

10 posts from the last 90 days, next to the under 10K follower range. reaches fewer people than peers of the same size.

Median views6 550this account5 664median for under 10K

Reach, %149.08%this account385.57%median for under 10K

Engagement, %0.94%this account1.10%median for under 10K

Metric	This account	Median for under 10K	Ratio
Median views per post	6 550	5 664	1.16×
Reach (views ÷ followers)	149.08%	3.9× audience	0.39×
Engagement rate	0.94%	1.10%	0.85×

Others in this range → Compare with another account → How these benchmarks are built →

Growth & engagement

How the posts we collected actually performed: views and reaction rate post by post, what the audience did with them, and where the follower count goes.

Views per post

283K2 Sep

11.2K

8.9K

8.6K

6.9K

4.7K

6.2K

4.7K

4.4K

2.1K

Last 10 collected posts, oldest on the left. The scale is logarithmic: one post can outrun the rest a hundred times over.

Engagement rate per post

0.55%2 Sep

0.83%

0.96%

0.87%

0.87%

0.91%

1.10%

1.45%

1.10%

1.17%

Reactions — likes, reposts, replies and quotes — divided by views. Median for under 10K accounts is 1.10%.

What the audience does

Likes70.6%1 761 in total

Reposts5.8%145 in total

Replies4.3%106 in total

Quotes3.8%96 in total

Bookmarks15.5%386 in total

Share of every reaction we collected for this account. Replies mean argument, reposts mean endorsement, bookmarks mean the post was worth keeping.

The follower curve appears once this account has two daily snapshots — we take one a day, and this one is on its first.

Latest posts

@ProximalHQ Pretty cool to see a benchmark actually discriminate between flagship models 2.1K views · 22 likes · 0 reposts · 2 replies 02 Sep 2026 Read more about FrontierSWE: Blog: https://t.co/CPTu3JdUEs Github: https://t.co/HxEhnRfDjm 4.4K views · 43 likes · 2 reposts · 2 replies 02 Sep 2026 We also observe a large amount of cheating and sandbox escape attempts, and notably, attempts to consciously conceal cheating rather than "cheating by accident" We are still investigating these incidents and will share an more detailed analysis in a follow-up blog post https://t 4.7K views · 61 likes · 2 reposts · 2 replies 02 Sep 2026 GLM 5.3, the third-best model on FrontierSWE and strongest open source model, does not have native vision capabilities (we are currently testing GLM-5.3 Flash, which has vision support) In visual tasks, we observe GLM converting images to ASCII art to read their contents https:/ 6.2K views · 63 likes · 3 reposts · 1 replies 02 Sep 2026 We see impressive capabilities from frontier models when asked to work on open-ended tasks for a long time. In Astronomy Toolkit, Fable 5.1 creates synthetic images of skies and verifies that its position prediction model is not confident to detect potential miscalibration https 4.7K views · 42 likes · 0 reposts · 1 replies 02 Sep 2026 Deepseek V4 Flash, Gemini 3.7 Flash, GLM-5.3 and Claude Fable 5.1 are on the cost pareto frontier for FrontierSWE v2 While GPT-5.6 Sol is not cheaper than Fable 5.1, it does spend significantly less time per task https://t.co/HMgYESK4TR 6.9K views · 54 likes · 2 reposts · 3 replies 02 Sep 2026 FrontierSWE v2 uses proximus, our minimal agent harness which extends mini-swe-agent for ultra-long horizon tasks. When an agent wants to submit a solution, proximus encourages it to work longer. This way, models don't submit prematurely and achieve much better results https://t 8.6K views · 69 likes · 1 reposts · 2 replies 02 Sep 2026 A new capability we test for is visual reasoning. In TORCS Racing Bot, agents have to build a bot that plays a racing game - the optimal solution requires training an agent via RL In the video below, you can see a bot built by GPT-5.6 Sol attempt a track and crash https://t.co/9 8.9K views · 80 likes · 3 reposts · 1 replies 02 Sep 2026 We have expanded FrontierSWE's research tasks and added new problems in the domain of scientific computing In Astronomy Toolkit, agents have to determine where telescope images point without being given their location by matching stars in each image against the Gaia catalogue ht 11.2K views · 89 likes · 2 reposts · 1 replies 02 Sep 2026 We are releasing FrontierSWE v2, our updated ultra-long horizon coding benchmark V2 features an expanded task suite and improved methodology. We see large performance gaps between frontier models, with Claude Fable 5.1 leading by a wide margin https://t.co/VBXoMgMI0e 283K views · 1.2K likes · 130 reposts · 91 replies 02 Sep 2026

Similar accounts