SShortSingh.
0
ProgrammingDEV Community ·

AI Agent Used 437,000 Tokens to Answer a Simple Question via MCP Server

Developers at FoxNose ran 18 real-world test scenarios connecting AI agents from Anthropic and OpenAI to their MCP server, uncovering severe token inefficiency in how agents handle tool calls. A single straightforward question triggered 20 consecutive tool calls, consuming 437,000 input tokens — because each new call re-sent the entire conversation history to the model. The core problem was not server errors or protocol failures, but bloated responses returning far more data than agents needed per call. The team addressed this by capping text field output at 1,000 characters with truncation markers and reducing the default page size to 5 results, cutting token usage in one scenario from 437,000 to 78,000. Their findings highlight that optimizing round-trip efficiency and response size matters far more than model selection when deploying AI agents in production.

0
IndiaTimes of India ·

Hockey India drops saffron jersey plan, reverts to blue for Asian Games

Hockey India has reversed its decision to outfit national teams in saffron jerseys, restoring the traditional blue kits for both the men's and women's squads. The change comes after the saffron jersey plan drew significant public and stakeholder criticism. The teams will wear blue as their primary kit at the upcoming Asian Games, where Olympic qualification is at stake. Saffron jerseys will, however, be retained as a secondary option if circumstances require.

0
IndiaNDTV ·

Houthis Destroy Saudi Weapons Trucks at Al-Wadi'ah; Riyadh Promises Retaliation

Yemen's Houthi forces targeted and destroyed Saudi weapons-laden trucks at the Al-Wadi'ah military camp. According to the Yemen News Agency, the Houthis conducted surveillance on Saudi camps and troops before carrying out the strike. The missiles used in the attack were reported to be locally manufactured by the Houthi movement. The Houthis identified a large concentration of vehicles carrying ammunition and weapons at the site prior to the assault. Saudi Arabia has vowed to respond to the attack following the destruction of its military assets.

0
IndiaNDTV ·

China Denies Arming Russia Despite Report of Drone Material Shipments

China has denied supplying weapons or military materials to Russia amid its ongoing war in Ukraine. However, a new report has uncovered documents suggesting shipments of polyacrylonitrile (PAN) precursor, a key material used in drone manufacturing, were sent from China to Russia. The material was reportedly listed under a customs code typically associated with civilian synthetic fibres like nylon and polyester, raising concerns about deliberate misclassification. The findings suggest that dual-use materials with military applications may be moving between the two countries under civilian trade labels. The report adds to growing international scrutiny over China's role in sustaining Russia's war effort.

0
ProgrammingDEV Community ·

How to Convert Excel Clock Times to Decimal Hours for Accurate Payroll

Excel timesheets record clock times, but payroll systems require durations expressed as decimal hours, making direct subtraction of start and end times insufficient. The core formula multiplies the time difference by 24, since Excel stores time as a fraction of a day, and any unpaid breaks must be subtracted before that conversion. A common error is treating 7 hours 30 minutes as 7.30 instead of the correct 7.50, since the decimal portion represents a fraction of an hour, not a minute count. Overnight shifts require a MOD function or a conditional day-addition to avoid negative results from standard subtraction. Once daily totals are converted to decimals, weekly sums and overtime calculations using standard Excel formulas become straightforward and payroll-accurate.

0
ProgrammingDEV Community ·

PostgreSQL Expands MultiXactOffset to 64 Bits, Eliminating Key Transaction Limit

PostgreSQL developer Heikki Linnakangas committed a patch widening the MultiXactOffset counter from 32 to 64 bits, effectively removing the previous cap of roughly 4 billion multitransaction members. The change eliminates MultiXactOffset wraparound, a condition that could previously force disruptive anti-wraparound vacuum operations on busy databases. Multitransactions are used internally by PostgreSQL whenever multiple transactions lock the same row, such as during SELECT FOR SHARE, FOR UPDATE operations, or foreign key enforcement. The patch author, who contributed directly to this work, notes that the new 64-bit limit is so large it is unlikely to be exhausted in any practical scenario. The commit has remained stable for nearly three months, signalling its acceptance into the PostgreSQL codebase.

0
IndiaTimes of India ·

Santa Cruz Island Feral Sheep Cut from 21,500 to Under 100 in 7-Year Cull

Conservationists carried out a large-scale removal of feral sheep from Santa Cruz Island to restore its fragile native ecosystem. Over seven years, the population was reduced from approximately 21,500 animals to fewer than 100. However, the effort risked wiping out a rare heritage breed with unique genetic traits. To prevent its extinction, twelve lambs were rescued and moved to the mainland to establish a captive breeding program. The initiative aims to preserve both the island's ecology and California's agricultural heritage.

0
IndiaTimes of India ·

Tim Cook says he never tried to emulate Steve Jobs as Apple CEO

In August 2011, Steve Jobs personally called Tim Cook to initiate a planned leadership transition at Apple, aware that his health was deteriorating. Jobs nominated Cook as his successor, a request that came as a surprise to Cook. Jobs expressed confidence in Cook's distinct abilities, acknowledging that his approach differed significantly from his own. Cook has since stated that rather than trying to replicate Jobs, he focused on being the best version of himself in the role.

0
ProgrammingDEV Community ·

Agent-Based Scanner Catches 95% of Malicious AI Skills vs. 8% for Static Tools

Researchers at Trent AI benchmarked five ClawHub skill scanners against a manually labelled set of 60 skills, covering benign, vulnerable, and malicious categories. Their agent-based scanner, trentclaw, achieved a 94.6% recall rate — over 40 percentage points higher than the next-best tool, ClawScan, which caught 54.1%. Signature and static analysis scanners performed worst, flagging as few as 8.1% of dangerous skills. The study found that reasoning-based scanners outperform pattern-matching tools partly because some malicious skills ship with no executable code, making them invisible to signature checkers. The benchmark comes amid real-world threats like the February 2026 ClawHavoc campaign, in which disguised productivity skills were used to steal API keys and browser credentials from users.

0
IndiaNDTV ·

Vijay's Jaw Gesture Sparks Political Row Over MK Stalin's 1976 Arrest Reference

Tamil actor-turned-politician Vijay is facing criticism from the ruling DMK and its ally CPM over a jaw gesture he made at a public event. Both parties have interpreted the gesture as a deliberate reference to Chief Minister MK Stalin's jaw, which was reportedly fractured during police custody. Stalin was allegedly assaulted after being arrested during the Emergency period in 1976. The CPM, a DMK alliance partner, has also condemned the gesture, making the controversy politically significant. Vijay has not yet issued an official response to the allegations.

0
IndiaTimes of India ·

Patrick Mahomes Builds Kansas City Sports Empire With Three Ownership Stakes

NFL star Patrick Mahomes has expanded his presence beyond football by becoming a part-owner of three professional sports franchises. His foray into sports ownership began with a reported $10 million investment in MLB's Kansas City Royals. He subsequently acquired ownership stakes in MLS club Sporting Kansas City and the NWSL's Kansas City Current, the latter alongside his wife Brittany. While still active as an NFL player, Mahomes is steadily building a significant business legacy rooted in Kansas City.

0
ProgrammingDEV Community ·

AI Coding Tools May Be Eroding Developer Understanding, Not Just Effort

A developer essay on DEV Community argues that AI-assisted coding has introduced two distinct forms of laziness: one that productively skips repetitive typing, and one that dangerously offloads genuine understanding. The author draws on Perl creator Larry Wall's definition of productive laziness — expending effort to reduce overall toil — to frame the first kind as acceptable and historically normal. The concern lies with the second type, where AI-generated code gets merged without any team member truly understanding or being able to fix it. Unlike copying from Stack Overflow, which required reading and adapting answers, AI suggestions appear pre-formatted and demand no active comprehension. The author proposes a self-check: could you recreate what the model wrote from scratch, even imperfectly, if the file disappeared?

0
IndiaTimes of India ·

Illegal promotion move costs Abhimanyu Mishra winning position against Mendonca in GCL

Indian chess prodigy Abhimanyu Mishra lost a seemingly certain win against Leon Luke Mendonca during a Global Chess League (GCL) encounter due to an illegal promotion sequence. With only seconds on his clock, Mishra accidentally dropped his rook while attempting the promotion, causing the move to be ruled incomplete and forcing him to redo it. The time lost in replaying the move left him with just one second remaining, and he flagged immediately after. The dramatic turn of events handed Ganges Grandmasters a match-winning result despite Mendonca reportedly apologising to Mishra for the unfortunate outcome.

0
ProgrammingDEV Community ·

Why Adding a New Feature to a Platform Costs Far More Than It Appears

A software platform essay explains that adding a new 'section type' — such as an events calendar — is far more complex than building a single screen. Every new content type must be understood by all platform systems, including navigation, design themes, push notifications, deep links, chatbots, and AI agents. The cost compounds over time because each new system added later must also learn every previously defined content type. The author contrasts this with bespoke, app-specific sections that skip platform-wide integration and remain cheap precisely because they teach nothing to the broader system. The piece argues that the true price of a new platform feature is not its launch effort but the permanent, growing maintenance burden it adds to the entire ecosystem.

← NewerPage 867 of 4703Older →