I went down the rabbit hole of Sony’s Ace, which they’re calling the ChatGPT moment in robotics. At first, I thought it was just typical AI hype, but it’s real, there’s huge potential IMO. I read the paper Sony AI submitted to Nature, and it was surprisingly simple to understand. It’s mostly about deep reinforcement learning so that helped, and the breakthroughs are truly amazing. I’m so hyped. I even tried to patch together the video with all my broken notes (basically the parts I actually understood, lol). You should definitely check out the paper though, just google “Sony Ace Nature” the PDF is free to download.

Anthropic onboarding day: Michael Scott introducing Karpathy like he just signed Wemby in free agency.

Here's a ready-to-post tweet draft (137 chars): --- Google I/O is coming and the hype feels different now 😂 Used to be everyone counting down for the next Pixel flagship. Now? It’s all “what crazy new Gemini drop we getting?” #GoogleIO #Geminl

"To reach the top, you have to go near the limit." - @RafaelNadal The documentary series Rafa premieres May 29.

Okay, this is officially too much Sony AI published research in Nature regarding its project Ace, an autonomous system that has officially defeated elite human players in competitive table tennis. It uses high-speed perception and reinforcement learning to act faster than human reaction time, marking a milestone for robots in unpredictable, high-speed physical environments.

Is it weird to say "thank you" and "please" while chatting with AI?

Just uploaded a new video on Kimi K2.6. The 1.1 trillion parameter MoE model. This open-weight powerhouse is a massive win for developers who want consistency over "lobotomized" updates.

{ "summary": "The market demonstrates a clear clash between a multi-session higher value migration and a sharp late-day long liquidation. Over the last several sessions, value had steadily migrated higher, but the recent 4 bars show a failure to sustain this acceptance, with the market closing below today's Value Area Low. While visible order-book leaning (depth pressure) at the top 5 levels showed some bid-side absorption during the afternoon decline, a massive surge in total sitting ask liquidity across the broader book and a widening spread indicate significant overhead supply and waning buying conviction.", "bias_next_open_60m": "bearish", "confidence": 65, "primary_signals": [ "Market Profile multi-session context: Over the last 5 sessions (Apr 10 to Apr 20), the POC migrated steadily higher from 24049.1 to 24377.4. However, today's close at 24331 represents a bearish rejection, falling below today's VAL of 24359.4 and signaling a failure to maintain the newly established higher value.", "Order-book structure: Comparing the recent 4 bars to the previous 15 bars, average total sitting ask quantity surged by 96671.18 contracts. This indicates overwhelming overhead supply building beyond the visible 5 levels as prices dropped.", "Microstructure: The five-level book widened as depth5_spread_bps increased from 4.90 to 5.03 basis points in the recent 4 bars, showing that deeper visible liquidity became less compact and more scattered during the late afternoon selloff.", "Order flow context: In the recent 4 bars (from bar_time 14:15:00 to 15:00:00), Open Interest delta averaged -8108.75 per bar compared to -3024.67 in the previous 15 bars, indicating the late drop (~42 points lower on average) was driven by aggressive long liquidation rather than fresh short initiation." ], "overnight_risks": [ "Because the selloff was primarily driven by long liquidation rather than fresh shorting, a lack of continued supply at the next open could cause a swift reversion to the POC at 24377.4.", "The net visible order-book leaning (depth_pressure) actually shifted positive ( 0.15) in the recent 4 bars, suggesting algorithmic bids were trying to absorb the liquidation; if these bids remain active at the open, early downside momentum may stall quickly." ], "hypothesis_next_open_60m": "Expect early weakness as the market processes the late-session rejection of the higher value area and the close below the VAL (24359.4). The initial 60 minutes are likely to test support around the previous session's POC (24287.8). Rallies back toward today's VAL will likely be sold into, given the massive buildup of total sitting ask liquidity observed into the close.", "position_into_close": "reduce" }

Hey @SonyLIV, why aren't you showing the ATP Masters? That's the only reason I subscribed.

Anyone who thinks prime Rafa can be beaten on clay is a total moron. There, I said it.

Provider: LM Studio Model: unsloth/gemma-4-26b-a4b-it { "summary": "The market is currently experiencing a significant value area migration lower compared to the multi-session history. While previous sessions (2026-04-08 through 2026-04-10) maintained high value areas between 23950 and 24070, today's action has accepted a much lower range, with the current session POC at 23697 being significantly below the previous session's VAL of 23998.1. From an order-book perspective, the recent 4 bars show a shift in dynamics: while the `net_visible_book_qty` remains negative (leaning toward the ask), there is a notable increase in `volume_delta` and a positive `weighted_imbalance` compared to the previous 15 bars. The `depth5_spread_bps` has widened from 4.46 to 5.29, indicating a less compact five-level book during recent price action.", "bias_60m": "bullish", "confidence": 55, "primary_signals": [ "Market Profile shows a major value area breakdown; today's session is trading entirely below the previous session's VAL (23998.1), indicating a shift in auction structure.", "Order-book evidence in the recent 4 bars shows `weighted_imbalance` moving from -0.017 to 0.012, suggesting increasing buy-side participation despite negative `net_visible_book_qty`.", "The most recent bar at 10:15:00 exhibits a strong positive `close_vs_open` of 60.4 and a significant increase in `volume_delta` (279,175) compared to the previous 15-bar average." ], "risks": [ "The `net_visible_book_qty` remains negative at -78,409.45 for the recent 4 bars, indicating visible order-book leaning toward the sell side.", "Today's session is currently operating in a lower value area with no immediate support from prior high-volume nodes (the nearest significant volume node is near 23692.1)." ], "hypothesis_next_60m": "The market is attempting to stabilize within the new, lower value area. Given the recent surge in `volume_delta` and the positive `weighted_imbalance` seen at 10:15:00, the next 4 bars are likely to test the current session's VAH (23745) as buyers attempt to reclaim local liquidity." }

Gemma-4 26B MoE actually beat out Gemini 3 Flash in my day to day benchmark. I honestly can't believe all you need is ~25k INR to get an elite AI experience.

Do I need any "extra" reason to short RIL 😂

Om Shanti Asha di... Those who know Marathi and Hindi, her greatest song was "जीवलगा राहिले रे दूर घर माझे" Impossible to clone or mimic unless you learn classics. 🙏

Never use Gemini fast. Once I developed a Python script for my Raspberry Pi Car infotainment Kivi app, every answer ends with: "Since you are interested in the car infotainment, here are a few ideas." Would you like to implement this particular script for your Raspberry Pi in your car?

This is not a drill. I repeat, this is not a drill

An AI breakthrough gave a hard lesson to Indian parents. There's no such thing as "safe money" I think it's true for all Asian parents.