Sathya (@sathya_ui_dev) header
Sathya (@sathya_ui_dev) profile picture

AI Architect building with GenAI & React โš™๏ธ Decoding tech, AI & Indian markets in plain English ๐Ÿ“ˆ Chennai | Not investment advice

Chennai
Joined May 2021
Photo by Sathya (@sathya_ui_dev)Photo by Sathya (@sathya_ui_dev)Photo by Sathya (@sathya_ui_dev)Photo by Sathya (@sathya_ui_dev)Photo by Sathya (@sathya_ui_dev)Photo by Sathya (@sathya_ui_dev)
A simple blinking cursor may be draining your battery. I tasked opus to debug an issue that the cpu was above my target of <0.1% cpu usage when an idle window of opencraft was open. And it showed that 1.5% of a cpu core was engaged just in blinking that cursor causing a window redraw.
7
2
1
29
1,409
Sathya retweeted
In 2016, a developer unpublished an 11-line JavaScript package called left-pad from npm after a trademark-related dispute over another package. within minutes, builds depending on babel, react-related tools, and thousands of other projects started failing because npm could no longer find the package. the disruption lasted 2.5 hours before npm restored the missing version. all because thousands of developers had come to depend on a tiny while loop that adds characters to the left of a string.
6
4
63
6,329
Sathya retweeted
This Anthropic tool use paper is f*cking gold Anthropic's own team took Opus from 49% to 74% accuracy and cut token use by 85%. And the model wasn't the upgrade. The upgrade: they stopped feeding Claude every tool and every raw result. Tool search โ†’ defer loading โ†’ code calls โ†’ parallel โ†’ examples The 6 parts of advanced tool use: Part 1 โ†’ Tool search 58 tools ate ~55,000 tokens before the agent read a word. Search loads ~500 tokens and finds the 3-5 it needs 77,000 โ†’ 8,700 tokens, and on their MCP evals Opus 4 went from 49% to 74% Part 2 โ†’ Defer loading Every tool marked defer_loading, only your 3-5 most used stay loaded Search "github" and only the GitHub tools show up Part 3 โ†’ Call tools from code One script reads 2,000+ expense lines in a sandbox The model only sees the 3 people over budget. 200 KB โ†’ 1 KB Part 4 โ†’ Run in parallel 20 tool calls in one code block, 19+ inference passes skipped 43,588 โ†’ 27,297 tokens Part 5 โ†’ Show examples The schema says "string". The example says "2024-11-06" Accuracy on complex parameters: 72% โ†’ 90% Part 6 โ†’ Fix the bottleneck Bloat โ†’ search. Big results โ†’ code. Bad params โ†’ examples Opus 4.5 went from 79.5% to 88.1% More tools in context made the agent dumber. Fewer tokens made it smarter. I packed all 6 parts into one prompt. Drop it into Claude Code and it sets the whole tool stack up for you. 6 parts. 1 prompt. 49% โ†’ 74%. Then read the article below โ†“
Your agent passed every test The issue is still open A better prompt won't fix it, because the model isn't the broken part I broke down the harness, loop and graph around it: what each one owns, how they nest, and which one to fix first โ†“
37
7
3
111
6,423
Tf ๐Ÿ˜‚๐Ÿ˜‚๐Ÿ˜‚
106
752
85
8,746
583,484
Gรผnรผn en tuhaf gifini buldum ๐Ÿ˜‚๐Ÿ˜‚
50
187
20
5,356
442,062
เฎคเฏ‹เฎณเฎฟเฎฒเฏ เฎšเฏเฎฎเฎจเฏเฎคเฏ เฎšเฏ†เฎฒเฏเฎฒเฏเฎฎเฏ CM Vijay๐Ÿฅน
6
560
14
2,690
92,336
๐Ÿ’ฐ India's Richest YouTubers ๐Ÿ‡ฎ๐Ÿ‡ณ 1. Tanmay Bhat โ€“ โ‚น665 Cr 2. Technical Guruji โ€“ โ‚น356 Cr 3. Samay Raina โ€“ โ‚น140 Cr 4. CarryMinati โ€“ โ‚น131 Cr 5. Bhuvan Bam โ€“ โ‚น122 Cr 6. Amit Bhadana โ€“ โ‚น80 Cr 7. Triggered Insaan โ€“ โ‚น65 Cr 8. Dhruv Rathee โ€“ โ‚น60 Cr 9. Ranveer Allahbadia โ€“ โ‚น58 Cr 10. Sourav Joshi โ€“ โ‚น50 Cr Who surprised you the most? ๐Ÿ‘‡
159
91
13
1,646
410,495
Running back in time with our timeless! ๐Ÿฅน๐Ÿ’› #ForOurThala
101
2,999
33
25,833
379,004
SHARP WORK FROM YASHASVI JAISWAL.
18
200
5
7,170
196,712
CUES FOR TODAY - Gift Nifty -150 pts lower, muted start - Nifty up 2nd day yesterday; Bank Nifty at record high - Wall St sold off, but Asia rallying on Nvidia beat - Brent ~$92/bbl, Rupee strengthens to 93.4 - FOMC Oct minutes tonight; Anthropic IPO pricing today WALL ST OVERNIGHT Dow -1.64% S&P -1.03%; 5-month low Nasdaq -0.74% Key Stocks that moved? Salesforce -9.4% Intuit -7.2% ServiceNow -6.5 Big After Hours Move - Nvidia +4.7% - Q3 rev $115.8bn, +103%; Q4 guide ~$130bn Asian Indices - Kospi +3%, SK Hynix +6%, Samsung +5% - Nikkei +1%, SoftBank +7% Cash Market Flows (17 Nov): DII +196.93 Cr FII +4,195.38 Cr F&O Cues FIIs 67% short, 33% Long Top Stocks To Track - IIFL Finance: Fairfax block deal, up to 13.25% stake - Cipla: USFDA clears InvaGen, New York with NAI - Groww: Ribbit sells 8.7cr shares worth 1,905 cr at 219 #StockMarket #CuesForToday @CNBCTV18Live
6
9
52
7,324
fun experiment: a drop-in API electron replacement with no JS runtime or chromium/webkit. all JS (including frontend) is compiled to native with Porffor. HTML/CSS rendered using @nicoburns' Blitz. here is unmodified GitHub Desktop, very buggy but works!
19
11
3
237
7,936
BREAKING: Nasdaq 100 futures extend gains to +1.3% in after hours trading in reaction to Nvidia's, $NVDA, earnings report.
123
147
34
2,340
227,828
Sathya retweeted
Mumbai Indians let go of Deepak Chahar and Trent Boult, which frees up INR 21.75 crore in their purse for the IPL 2027 auction srkl.in/6014BCSyBY
8
38
16
736
62,925
SOUND ON ๐Ÿ”Š The name is Dhoni Mentor Dhoni, 007 Reloaded #ForOurThala
350
9,318
449
45,198
1,888,933
Sathya retweeted
3 things to check before blaming the ai for broken code: 1. did you give it the actual error, or just "it's not working" 2. did you tell it which versions you're on 3. did you run what it wrote, or just read it and nod
13
2
15
535
๐Ÿšจ WHAT ACTUALLY HAPPENS WHEN YOU CLICK "BUY NOW"? JUNIOR ENGINEER: -Call the payment API. -Update the database. -Show success. SENIOR ENGINEER:
3
1
20
18,040
Sathya retweeted
the model you default to is rarely the best one for the job. it's just the one you already had open. half my prompts would run fine on something way cheaper if i actually bothered to switch
3
1
2
167
Nobody wants to be a beginner anymore. ChatGPT gives you answers. Claude explains everything. Codex and Cursor help you build. But nobody becomes an expert without being bad at something first. AI is making things easier, but the mistakes and struggles are part of what teach us. @grok if AI makes everything easier, what happens to the skills we used to learn the hard way?
8
2
9
357